Maximum High Jump with a Robotic Leg - Student Projects
Maximum High Jump with a Robotic Leg - Student Projects
Maximum High Jump with a Robotic Leg - Student Projects
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Master Thesis<br />
<strong>Maximum</strong> <strong>High</strong> <strong>Jump</strong> <strong>with</strong> a<br />
<strong>Robotic</strong> <strong>Leg</strong><br />
April 2008<br />
Supervised by: Author:<br />
C. David Remy Jonas Fisler<br />
Prof. Keith Buffinton<br />
Autonomous Systems Lab<br />
Prof. Roland Siegwart
Abstract i<br />
Abstract<br />
<strong>Jump</strong>ing is one of the most dynamic movements that animals and humans can<br />
perform. It is therefore no surprise, that jumping poses many challenges to<br />
a robotic system. A leg that is able to jump requires a durable mechanical<br />
construction, powerful actuation, and sophisticated control algorithms.<br />
In this thesis a three degree of freedom robotic leg was developed that is<br />
capable of such a highly dynamic movement. This included a sophisticated<br />
mechanical construction that minimized the overall mass and especially the<br />
inertia of the leg, series elastic actuation to recover energy, increase the peak<br />
power of the motors and enhance the shock tolerance of the actuators, and<br />
control inputs that optimally exploited the stretch-shortening-cycle of these<br />
actuators.<br />
Within the project, several leg design variations were evaluated. The most<br />
promising ones were then modeled and an optimal combination of motor, gear,<br />
and elasticity was determined in simulation. The final mechanical design was<br />
performed in CAD, resulting in production ready drawings of all parts. After<br />
successful assembly, the capability of the leg to jump was shown in experiments.<br />
Furthermore, an optimal feedforward control algorithm was proposed which can<br />
significantly improve the jumping performance of the leg.<br />
The results of this thesis are intended to be used in the develop of a<br />
quadruped robot that is capable of dynamic walking and running.
Contents ii<br />
Contents<br />
Abstract i<br />
List of Tables iii<br />
List of Figures iii<br />
1 Introduction 1<br />
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1<br />
1.2 Previous work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1<br />
1.3 Objectives of this thesis . . . . . . . . . . . . . . . . . . . . . . . 3<br />
1.4 Organization of this thesis . . . . . . . . . . . . . . . . . . . . . . 4<br />
2 Design Evaluation and Simulation 5<br />
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5<br />
2.2 First design evaluation and kinematic models . . . . . . . . . . . 5<br />
2.3 Dynamic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10<br />
2.4 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13<br />
3 Mechanical Design 20<br />
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20<br />
3.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20<br />
3.3 Spring selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20<br />
3.4 Gear selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21<br />
3.5 Hip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24<br />
3.6 Thigh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24<br />
3.7 Knee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25<br />
3.8 Shank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />
3.9 Cable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />
3.10 Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />
4 Control 30<br />
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />
4.2 Improved model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />
4.3 Optimized feedforward control . . . . . . . . . . . . . . . . . . . 33<br />
4.4 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />
5 Assembly and Test 39<br />
5.1 Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39<br />
5.2 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
List of Tables iii<br />
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43<br />
6 Summary and Future Work 47<br />
6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47<br />
6.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47<br />
A Appendix 48<br />
A.1 Derivation of kinematic equations for the pantographic design . . 48<br />
A.2 Morphological box . . . . . . . . . . . . . . . . . . . . . . . . . . 49<br />
A.3 Data sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53<br />
A.4 Parts list . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62<br />
References 63<br />
List of Tables<br />
1 Design considerations . . . . . . . . . . . . . . . . . . . . . . . . 11<br />
2 Some values comparing final velocity to jump height . . . . . . . 14<br />
3 Overview motors and gearboxes . . . . . . . . . . . . . . . . . . . 15<br />
4 Optimized jump height . . . . . . . . . . . . . . . . . . . . . . . . 17<br />
5 Comparison of three compression springs . . . . . . . . . . . . . . 21<br />
6 Mass properties of the leg identified from the 3D-CAD model . . 32<br />
List of Figures<br />
1 Series elastic actuation . . . . . . . . . . . . . . . . . . . . . . . . 2<br />
2 Previous leg designs . . . . . . . . . . . . . . . . . . . . . . . . . 3<br />
3 Three designs by Toepfer . . . . . . . . . . . . . . . . . . . . . . 4<br />
4 Mechanical design variants . . . . . . . . . . . . . . . . . . . . . . 5<br />
5 Consideration to motor torque . . . . . . . . . . . . . . . . . . . 6<br />
6 Pantographic leg design . . . . . . . . . . . . . . . . . . . . . . . 7<br />
7 Ballpoint leg design . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
8 Motor knee design . . . . . . . . . . . . . . . . . . . . . . . . . . 9<br />
9 Cable design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9<br />
10 Comparison of ∂y<br />
∂φ . . . . . . . . . . . . . . . . . . . . . . . . . . . 10<br />
11 Block diagram of dynamic model . . . . . . . . . . . . . . . . . . 10<br />
12 Simplified leg in SimMechanics . . . . . . . . . . . . . . . . . . . 13<br />
13 Comparison of motor optimization . . . . . . . . . . . . . . . . . 16<br />
14 Optimized launch for a jump . . . . . . . . . . . . . . . . . . . . 17
List of Figures iv<br />
15 Force on the ground before take-off . . . . . . . . . . . . . . . . . 18<br />
16 Differential drive . . . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />
17 Load estimation hip . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />
18 Hip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23<br />
19 Hip detail 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24<br />
20 Hip detail 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25<br />
21 Hip detail 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />
22 Thigh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27<br />
23 Knee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28<br />
24 Cable winding knee . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />
25 Shank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />
26 Dyneema cable . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />
27 <strong>Leg</strong> geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31<br />
28 SimMechanics model . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />
29 Power contraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34<br />
30 Trajectories 1 of optimized jump . . . . . . . . . . . . . . . . . . 37<br />
31 Trajectories 2 of optimized jump . . . . . . . . . . . . . . . . . . 38<br />
32 Circlip of the knee shaft . . . . . . . . . . . . . . . . . . . . . . . 39<br />
33 Mounting torsion spring . . . . . . . . . . . . . . . . . . . . . . . 40<br />
34 Support of RE 25 . . . . . . . . . . . . . . . . . . . . . . . . . . . 40<br />
35 Cable on pulley . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41<br />
36 Test setup: mounting . . . . . . . . . . . . . . . . . . . . . . . . . 42<br />
37 Trajectory control . . . . . . . . . . . . . . . . . . . . . . . . . . 45<br />
38 Picture series of a jump . . . . . . . . . . . . . . . . . . . . . . . 46
1 Introduction 1<br />
1 Introduction<br />
1.1 Motivation<br />
Most commonly, machines move around on wheels because they can achieve very<br />
good efficiencies <strong>with</strong> relatively simple mechanical implementations. Balance<br />
is usually not an issue for wheeled systems, and they can move at high speed.<br />
But these amenities are limited to flat surfaces.<br />
On rough terrain legged systems have clearly advantages. They can climb<br />
up mountains and stairs, wade in mud or snow, or balance over rocks. When<br />
walking, small obstacles can be overstepped, and the direction of travel imme-<br />
diately changed. If the system stumbles and falls it can rise again <strong>with</strong> the<br />
help of legs. Another advantage of legged systems is their ability to manipulate<br />
objects like for example a ball.<br />
Engineers gained interest to build artificial legs to overcome the drawbacks<br />
of wheels. A recently launched project at the Autonomous Systems Lab at<br />
ETH Zurich deals <strong>with</strong> the design and construction of an energy efficient robotic<br />
quadruped that is capable of various dynamic motions including fast walking,<br />
trotting, bounding, and galloping. The robot will exploit natural passive dy-<br />
namics to be energetically efficient while maximizing dynamic performance.<br />
Within this context, there are two main difficulties. Firstly, the various gaits<br />
show complex patterns, which pose difficult challenges on the computational<br />
algorithms and the computational power. Secondly, rapid leg movements or<br />
jumps need explosive leg power. Power is usually heavy to carry around and<br />
poses a major limitation.<br />
Keeping a dynamically moving robot stable requires a sophisticated control<br />
algorithm, which necessitates quick and accurate actuators. To allow a robot<br />
to run fast and efficient it must be able to place its legs in the short instance<br />
between two steps and -in about the same time- deliver the necessary impulses<br />
to keep its center of gravity in the air.<br />
1.2 Previous work<br />
1.2.1 <strong>Leg</strong>ged robotics<br />
The first four-legged autonomous robot, the “phoney pony“, walked in 1966 [1,<br />
2]. Matsuoka [3] realized a monopod able to move dynamically, and Raibert et<br />
al. [4] developed the first dynamically walking quadruped. In recent years,<br />
efforts increased to develop legged robots capable of dynamic maneuvers like<br />
jumping, bounding, trotting, and galloping [5]-[14].
1.2 Previous work 2<br />
To build legged systems, Robinson et al.recommend to use series elastic<br />
actuation, because this introduces serious advantages for electric actuators.<br />
Figure 1: Series elastic actuation<br />
Figure 1 demonstrates the basic principle of series elastic actuation. The torque<br />
or force of the motor is transmitted to the load through a spring. This decouples<br />
the actuators position from the load position, and low-pass filters the actuator’s<br />
output, leading to the following advantages.<br />
Shock tolerance of the actuator is increased. It has a lower reflected inertia,<br />
and causes less damage to the environment. More stable and accurate force<br />
control is possible. Force band<strong>with</strong> (especially for spring- or damper-like out-<br />
put) is increased except for zero motion. The elasticity additionally implements<br />
a means to store energy, particularly for systems that undergo periodic motion.<br />
Series elastic actuation should be used when high constraints are posed onto<br />
the power to mass ratio, or impact shocks might damage actuator and gears.<br />
For a careful selection process of the components for series elastic actuation,<br />
Robinson et al. [15] show some design guidelines. In 2006, Paluska et al. [16]<br />
investigate actuator power and work output of electric motors and state that<br />
series elastic actuation can temporarily increase the motor energy delivered to<br />
a mass by a factor of four.<br />
1.2.2 <strong>Leg</strong> design for dynamic motion of a quadruped<br />
In 1995, Mennitto [17] (Fig. 2(a)) developed a 8 kg three limb leg prototype<br />
for a quadruped. The actuation is built on a concept called LADD (linear<br />
to angular displacement device). The LADD is very efficient and light-weight<br />
but there is no commercially available solution, and the manufacturing of the<br />
LADD is complex and takes a long time to test.<br />
Remic [18] built a protoype leg, in 2005, consisting of two limbs (Fig. 2(b)).<br />
Weaknesses of this design were corrected and a theoretical model built by<br />
Knox [19] and Curran [20]. The leg is able to successfully perform a series<br />
of vertical jumps. Their work has a similar goal as the project at ETH, but<br />
does not show the evaluation and selection process of the leg construction.<br />
In a thesis on legged robotics at the Autonomous Systems Lab, Toepfer [21]<br />
evaluated three leg designs: ’Linear leg’, ’2-motor-leg’ <strong>with</strong> linear elasticity or
1.3 Objectives of this thesis 3<br />
(a) Mennitto (b) Remic<br />
Figure 2: Previous leg designs by Mennitto and Remic<br />
<strong>with</strong> rotational elasticity, and ’pantographic leg’ (Fig. 3). After simulation, he<br />
concluded that the design ’linear leg’ is inferior to the other designs, and rec-<br />
ommended the ’pantographic leg’, of which a first CAD-model was developed.<br />
1.3 Objectives of this thesis<br />
In this thesis a single prototype leg is developed meeting the requirements for<br />
dynamic movements. It is tested in a scenario consisting of a single jump <strong>with</strong><br />
the overall goal to maximize a high jump. Experience shall be gained for further<br />
improvements and construction of robotic legs.<br />
The challenge in the development of the mechanical system lies in the nec-<br />
essary dynamics. The moving parts are required to have small masses and<br />
moments of inertia, and must yet be able to deliver large forces and moments.<br />
The design has to <strong>with</strong>stand high loads during landing and take-off. Moreover,<br />
it is desired that the structure is able to passively store energy and recover it at<br />
a later instance, which significantly improves the efficiency of the locomotion.<br />
Previous studies at the lab estimated a mass of 5 kg for the leg including<br />
a fourth of payload like cameras or sensors. The leg length should be ap-<br />
proximately 40 cm long consisting of two limbs. Although for a jump the full<br />
movement range is not necessary, the leg is supposed to be able to fold and to<br />
stretch as far as possible. This will allow future experiments like for example<br />
standing up from the ground.<br />
Electrical energy supply is limited by a maximum voltage of 36 V, providing
1.4 Organization of this thesis 4<br />
(a) Linear leg (b) 2-motor-leg (c) Pantographic<br />
leg<br />
Figure 3: Three designs that were evaluated by Toepfer [21].<br />
a maximum current of 10 A.<br />
1.4 Organization of this thesis<br />
In Chapter 2, studies on several designs are modeled and simulated to analyze<br />
the occuring forces and moments before take-off. An optimization algorithm<br />
is run to determine a good selection of motor, gearbox and elasticity to reach<br />
maximal height during a jump<br />
The detailed design of the leg, and the evaluation process of the mechanical<br />
parts are explained in Chapter 3.<br />
In Chapter 4, the modeling is revised and adapted to the actual leg construc-<br />
tion. An optimal feedforward control trajectory for the motors is developed and<br />
implemented into software.<br />
Section 5 finally shows the manufacturing and assembly of the prototype<br />
and the results and experiences that were made in a first series of tests.
2 Design Evaluation and Simulation 5<br />
2 Design Evaluation and Simulation<br />
2.1 Introduction<br />
In this section, four mechanical designs are analyzed and compared. The models<br />
for actuation (motor), elasticity (spring), and the mechanical dynamics of the<br />
selected design will be derived and simulated. Motor, spring, and leg are form-<br />
ing a series elastic actuator as described in section 1.2.1. With these models,<br />
optimization algorithms are run to select an optimal motor, spring and initial<br />
leg position to perform a maximal jump for each design. These components<br />
and the estimated forces and torques will form the basis for the mechanical<br />
construction in Chapter 3.<br />
2.2 First design evaluation and kinematic models<br />
Toepfer [21] compared several designs (Section 1.2.2), and proposed the ’pan-<br />
tographic leg design’ to achieve a maximum jump. Also, he pointed out that a<br />
’telescopic leg design’ is inferior to a leg <strong>with</strong> ’thigh’ and ’shank’.<br />
The main issue is to find a constructive solution on how to apply a torque to<br />
the knee while keeping the leg inertia small. A reevaluation of Toepfer’s models<br />
(except the ’telescopic leg’) was done and extended by two additional designs<br />
(Fig. 4).<br />
Figure 4: Mechanical design variants<br />
In future, we will refer to them as<br />
• ballscrew design<br />
• pantographic design<br />
• motor knee design
2.2 First design evaluation and kinematic models 6<br />
• cable design<br />
It should be noticed that between the motor and leg an elastic element has to<br />
be introduced to create series elastic actuation, which needs space, and might<br />
add constraints on the construction.<br />
Each design has a characteristic motor to end effector force relation Tm<br />
. It is Fg<br />
not only relevant how much torque Tm the motor can provide but also how much<br />
force Fg can be delivered to the ground (Fig. 5). This ratio is characteristic<br />
for each design depending on its kinematics, and is calculated using the power<br />
balance.<br />
Pin = Pout<br />
(2.1)<br />
Figure 5: Relevant for a high jump is actually not the motor torque but the<br />
delivered force to the ground<br />
2.2.1 Pantographic design<br />
The pantographic design (Fig. 6) consists of four bars that are linked <strong>with</strong> each<br />
other to a framework. The length of the thigh ( ¯ AB) and the shank ( ¯ BE) is l,<br />
the extension of the shank measures d. If the bar ¯ AD is fixed and is twice as<br />
long as the distance ¯ BC, the point E follows quite closely a straight trajectory<br />
(Fig. 6), when the bar thigh ( ¯ AB) is rotated by φ around A.<br />
With<br />
The power balance derives the following torque to force ratio<br />
the following equation is found<br />
Pin = T · ω = T dφ<br />
dt = Pout = F · v = F dy<br />
dt<br />
dy<br />
dt<br />
T<br />
F<br />
∂y ∂φ<br />
= ·<br />
∂φ ∂t<br />
= v<br />
ω<br />
(2.2)<br />
(2.3)<br />
∂y<br />
= . (2.4)<br />
∂φ
2.2 First design evaluation and kinematic models 7<br />
<strong>with</strong><br />
∂y<br />
∂φ<br />
(a) Pantographic<br />
leg design<br />
(b) Trajectory of pantographic leg<br />
Figure 6: Pantographic leg design<br />
The kinematic equation of the pantographic design E(φ) is (App. A.1)<br />
y(φ) = yC −<br />
l + d<br />
d<br />
· |−l cos φ − yC| (2.5)<br />
yC = −2d cos φ2 + a<br />
p (−l cos φ + 2d cos φ2) + h<br />
p (l sin φ − 2d sin φ2)<br />
�<br />
(2.6)<br />
p = (xB − xD) 2 + (yB − yD) 2<br />
(2.7)<br />
a = l2 − d 2 + p 2<br />
2p<br />
(2.8)<br />
h = � l 2 − a 2 (2.9)<br />
was derived numerically from the kinematic relationship.<br />
2.2.2 Ballscrew design<br />
Characteristic for the ballscrew design (Fig. 7) is a ballscrew motor along the<br />
thigh. The motor pulls and pushes on the shank extension to rotate the shank<br />
around the knee joint The actuator can rotate freely around the extension of<br />
the shank and the hip.
2.2 First design evaluation and kinematic models 8<br />
The power balance becomes<br />
Figure 7: Ballscrew leg design<br />
Fm ˙x1 = Fg ˙y (2.10)<br />
∂x1 ∂φ<br />
Fm<br />
∂φ ∂t<br />
Fm<br />
Fg<br />
∂y ∂φ<br />
= Fg<br />
∂φ ∂t<br />
= ∂y<br />
∂x1<br />
and the kinematic equations for the ballscrew design are<br />
Again, ∂y<br />
∂x1<br />
x1(φ) =<br />
�<br />
(l + d cos (π − 2φ)) 2 + (d sin (π − 2φ)) 2<br />
(2.11)<br />
(2.12)<br />
(2.13)<br />
y(φ) = −2l cos φ (2.14)<br />
is calculated numerically.<br />
2.2.3 Motor knee design<br />
In the motor knee design (Fig. 8), the motor is placed in the thigh close to the<br />
knee, so that the torque can be applied directly at the knee joint, for example<br />
<strong>with</strong> bevel gears. For this design, the power balance and the kinematic equation<br />
are<br />
∂y<br />
Tm ˙α = Fg ˙y = Fg ˙α (2.15)<br />
∂α<br />
y(φ) = −2l cos φ = −2l cos α<br />
(2.16)<br />
2<br />
The torque to force relationship can be easily calculated analytically.<br />
∂y α<br />
= l sin = l sin φ (2.17)<br />
∂α 2
2.2 First design evaluation and kinematic models 9<br />
2.2.4 Cable design<br />
Figure 8: Motor knee design<br />
The motor is placed at the hip and <strong>with</strong> a cable a force or torque is transmitted<br />
to the knee. This design has the same kinematic equation as the motor knee<br />
design if the torque is transmitted using pulleys (Fig. 9(a)) or as the ballscrew<br />
design if the cable is attached at two fixed points (Fig. 9(b)), so that for these<br />
designs the torque to force ratio does not need to be evaluated separately.<br />
(a) Cable design:<br />
Pulley<br />
(b) Cable design<br />
2: Fixed<br />
Figure 9: In the ’cable design’ the torque is provided at the hip and is applied<br />
to the knee via a cable. In design (a) the cable is tightend to a pulley producing<br />
a torque at the knee, whereas in design (b) the cable is fixed to the shank.<br />
2.2.5 Design evaluation<br />
Figure 10 compares the three kinematic results. Explicitely, they mean if the<br />
motor provides a constant torque T the force in y-direction F at the end effector<br />
is small where ∂y<br />
∂φ is large. Or considering that a constant force is desired, T<br />
must increase if ∂y<br />
∂φ increases.<br />
It is evident that the designs ’knee motor’ and ’cable motor’ <strong>with</strong> pulleys
2.3 Dynamic model 10<br />
have the best torque to force transmission (dotted curve). The pantographic<br />
design needs a doubled motor torque for a required torque at the knee due to<br />
the framework ratio ¯<br />
AD ¯<br />
BC . The ballscrew design shows a fastly rising curve that<br />
goes towards infinity if the angle φ gets close to 90 ◦ .<br />
It can be seen also that the pantographic and the ballscrew design are both<br />
limited in the angular range due to their construction.<br />
Figure 10: Comparison of ∂y<br />
∂φ<br />
Table 1 compares other properties of the design variants. As there are no<br />
clear advantage of the ballscrew or the pantographic design, the decision falls<br />
either towards the cable or the motor knee design, as a result of the comparsion<br />
of the torque analysis. They both have the same characteristic curve.<br />
2.3 Dynamic model<br />
Figure 11: Block diagram that shows the interdependence of the actuation<br />
and the leg<br />
The time-variant behavior of the system consisting of series elastic actuators<br />
and the robotic leg is simulated <strong>with</strong> a dynamic model.<br />
It consists of three parts. In the first part the motor is modeled. The second<br />
block contains the equations for the elasticity, and thirdly, in SimMechanics, a
2.3 Dynamic model 11<br />
Weight<br />
distribution<br />
Ballscrew<br />
design<br />
Knee motor in<br />
thigh, spindle<br />
of steel<br />
Transmission Torque at the<br />
knee doubles<br />
through the<br />
framework to<br />
the hip<br />
General remarks<br />
Range<br />
angle<br />
(Fig 6, 7, 8)<br />
Buckling of<br />
spindle possible<br />
Pantographic<br />
design<br />
Knee motor<br />
at the hip,<br />
aluminium<br />
framework at<br />
thigh<br />
Direct torque<br />
application<br />
possible<br />
Thigh and<br />
shank motion is<br />
coupled<br />
Motor knee<br />
design<br />
Knee motor at<br />
the knee<br />
Linear<br />
transmission<br />
Torque to the<br />
shank independent<br />
of thigh<br />
position<br />
Cable design<br />
Knee motor at<br />
the hip, pulleys<br />
necessary<br />
Pulleys: See<br />
motor knee<br />
design<br />
Fixed: See<br />
Ballscrew<br />
design<br />
Cable might be<br />
elastic<br />
≈10 ◦ - 80 ◦ 23 ◦ - 78 ◦ 0 ◦ - 90 ◦ See: Ballscrew<br />
and motor knee<br />
Table 1: Design considerations<br />
Simulink toolbox, the leg <strong>with</strong> the three bodies ’hip’, ’thigh’, and ’shank’ was<br />
built (Fig. 11).<br />
Input to the model is the motor voltage U. Depending on the angular<br />
position of the gearbox φgb and the position of the leg φl the torque of the<br />
elasticity Ts changes. The torque itself effects motor and leg. The velocity ˙y of<br />
the body can be read from the model as output.<br />
2.3.1 Motor<br />
In this project, DC motors are used for actuating the three degrees of freedom of<br />
the leg. Brush-type DC motors have a permanent-magnet stator so that a static<br />
magnetic field excitation is generated. Through this setup a back electromotive<br />
force (back emf) is induced in the rotor.<br />
The electric circuit inside the motor is described by<br />
Ua(t) = La · d<br />
dt Ia(t) + Ra · Ia(t) + Ui(t), (2.18)<br />
<strong>with</strong> the armature current Ia(t), the resistance Ra, the inductance La, and the<br />
voltage Ua(t). Ui(t) is the induced back emf.<br />
The induced voltage increases proportionally <strong>with</strong> the rotational speed ω,
2.3 Dynamic model 12<br />
and can be expressed as<br />
The torque of the motor is given by<br />
Ui(t) = κi · ω (2.19)<br />
Tm(t) = κa · Ia(t) (2.20)<br />
From these to equations it can be followed that the current decreases <strong>with</strong> a<br />
higher back emf, which has the effect of lowering the maximal torque.<br />
The speed constant κi and the torque constant κa are in theory equal.<br />
Rotational and other losses though are usually multiplied to the torque constant<br />
resulting in a lower κa than κi.<br />
According to Newton’s law, the motor torque Tm(t) acts on the motor shaft<br />
(acceleration and deceleration) and the spring Ts(t)<br />
Tm(t) = Jm ˙ω + Ts(t) (2.21)<br />
<strong>with</strong> Jm being the moment of inertia of the motor load (i.e. gearbox and half<br />
of the spring).<br />
The gearbox multiplies the torque by the gear ratio γ, but introduces friction<br />
losses ηgb.<br />
2.3.2 Elasticity<br />
Tgb = ηgbTmγ (2.22)<br />
To achieve momentarily maximum torque, a spring is inserted between the<br />
motor <strong>with</strong> a gearbox and the actuated body, as described in section 1.1. The<br />
elasticity obeys the linear relationship<br />
Ts = ks (φleg − φgb) . (2.23)<br />
The spring constant is ks. φleg − φgb defines the compression of the elasticity.<br />
2.3.3 <strong>Leg</strong> in SimMechanics<br />
The origin of the world coordinate frame lies at the contact point of the shank<br />
to the ground. The contact point was modeled as a revolute joint around<br />
the z-axis. The other degrees of freedom were locked. Shank and thigh were<br />
introduced as two beams <strong>with</strong> also a revolute joint in between. The hip was<br />
modeled as a sphere of mass mh at the end of the thigh that can slide along the<br />
vertical axis only (prismatic joint) and is connected to the thigh by a revolute
2.4 Optimization 13<br />
joint. The effect of a motor mass at the knee was analyzed by adding another<br />
sphere of mass mk there.<br />
Figure 12: Simplified leg in SimMechanics<br />
The inertia Js,t of the two beams (shank and thigh), and Jh,k of the spheres<br />
were defined by<br />
Js,t =<br />
Jh,k =<br />
⎛<br />
⎜<br />
⎝<br />
⎛<br />
⎜<br />
⎝<br />
r<br />
ms,t<br />
2 s,t<br />
2 0 0<br />
0 ms,t · (3r2 s,t +l2 )<br />
12<br />
0<br />
0 0 ms,t · (3r2 s,t +l2 )<br />
12<br />
mh,kr 3 0 0<br />
0 mh,kr 3 0<br />
0 0 mh,kr 3<br />
⎞<br />
⎞<br />
⎟<br />
⎠<br />
(2.24)<br />
⎟<br />
⎠ (2.25)<br />
where m is the mass of the body, l is the length, and r is the radius.<br />
2.4 Optimization<br />
2.4.1 Parameter<br />
The jump height was the criteria for choosing the optimal design. The jump<br />
height depends only on the final velocity at the moment of take-off and can be<br />
calculated by the equation for energy conservation. The change in the potential<br />
energy at take-off to the maximum height must be equal to the change in the<br />
kinetic energy.<br />
∆Ekin = 1<br />
2 mv2 (take − off) = mg∆h = ∆Epot<br />
h(v) = v2<br />
2g<br />
(2.26)<br />
(2.27)
2.4 Optimization 14<br />
A few values in tab. 2 give an impression of the necessary velocity to achieve<br />
a certain height.<br />
v(take − off) [m/s] 1 1.4 2.0 2.4 2.8 3.15 3.45<br />
h [cm] 5 10 20 30 40 50 60<br />
Table 2: Some values comparing final velocity to jump height<br />
It is therefore important that the final velocity before take-off is as large as<br />
possible. Take-off occurs, when the vertical force on the ground becomes zero.<br />
The last consideration that has to be made is that at take-off not every<br />
part of the leg has the same velocity. Thus, slower parts will be accelerated,<br />
while the hip mass is slowed down. This can be estimated <strong>with</strong> the equation<br />
for conservation of momentum.<br />
I = mtotvtot = mhipvhip + mlegvleg<br />
(2.28)<br />
Assuming a relatively homogeneous weight distribution in the thigh and the<br />
shank, the average vertical velocity of the leg vleg can be approximated to be at<br />
the middle of the leg at the knee and is then 0.5 · vhip. vtot is then approximated<br />
<strong>with</strong> mtot = mhip + mleg as<br />
vtot = vhip<br />
�<br />
1 − mleg<br />
�<br />
2mtot<br />
(2.29)<br />
indicating that a larger mass at the leg introduces velocity losses at take-off.<br />
Still, both options <strong>with</strong>out and <strong>with</strong> larger mass at the knee (motor knee design)<br />
will be optimized for comparison.<br />
At this point, the main challenge was the selection of suitable motors.<br />
Smaller motors have higher dynamic but the maximum torque is limited. Larger<br />
motors in turn provide more torque but become slow due to the high inertia<br />
and the armature resistance.<br />
Many parameters of the model and of the jump can be optimized: motor,<br />
time-varying voltage input, gearbox parameter, spring, mass distribution in<br />
the robot leg, and so on. To simplify the algorithm only three parameters were<br />
chosen to be optimized: the transmission ratio of the gearbox γ, the linear<br />
spring constant kspring k, and the initial leg position defined by the hip angle<br />
φinit.<br />
The main goal is to maximize the height of a jump, respectively the final<br />
vertical velocity. The cost functional in the optimization<br />
J = −1.5v 2 y hip (end) (2.30)
2.4 Optimization 15<br />
gives good results.<br />
The optimization was run for the six maxon motors given in tab. 3. DC<br />
motors from Maxon were chosen because they offer very high quality, are easily<br />
controllable, have a fast response, and also because the Autonomous Systems<br />
Lab has already lots of experience <strong>with</strong> them. The motor parameter can be<br />
found in the Maxon catalogue (App. A.3) and were inserted into to the motor<br />
model derived in section 2.3.1. Only six motors were selected because larger<br />
motors are too heavy whereas the smaller ones cannot be combined <strong>with</strong> gear-<br />
boxes that can hold moments above 10 Nm, which were required too for stat-<br />
ically holding the robots‘ mass. 7.5 Nm torque output of the gearbox means<br />
that there is an additional transmission level of at least 1:1.5 necessary that<br />
can be built implemented <strong>with</strong> bevel gears or pulleys.<br />
The maximum current was set to 10 A, which is the maximum that can be<br />
supplied by the power supply.<br />
Label Gearbox Maximal Motor Mass [g] Nominal<br />
torque [Nm] voltage [V]<br />
1 GP 52 C 22.5-45 RE 40 1100-1250 24<br />
2 GP 42 C 11.3-22.5 RE 36 710-810 32<br />
3 RE 35 700-800 30<br />
2 GP 32 C 7.5 RE 36 544 32<br />
3 RE 35 534 30<br />
4 RE 30 432 36<br />
5 RE 26 344 36<br />
6 RE 25 324 30<br />
Table 3: Overview motors and gearboxes<br />
At the initial leg position, the elasticity is assumed to be relaxed so that at<br />
start the leg will fold until the motor and spring torque hold it and then extend<br />
the leg again, accelerating the hip upwards. To reduce optimization parameters<br />
the knee motor is assumed to be running <strong>with</strong> full torque at all time, keeping<br />
in mind that the maximum torque changes depending on the motor velocity<br />
(Sec. 2.3.1). The simulation is stopped when the vertical force on the ground<br />
becomes positive, and thus the leg takes off from the ground. The velocity of<br />
the hip at this instant, is inserted into eq. 2.29 to determine the jump height.<br />
A simulation cycle is not only stopped, when either the force to the ground<br />
becomes positive (take-off criterion), but also when the hip comes closer to the<br />
ground than 5 cm (crash criterion), or when the leg is fully stretched (maximal<br />
range criterion). Because the cost functional (Eq. 2.30) is certainly largest if<br />
the simulation stops at the take-off criterion the optimization finds solutions
2.4 Optimization 16<br />
for a maximal jump.<br />
It was also investigated if there was an influence on the height of a jump if<br />
the motor is placed at the knee (direct application of torque to the knee joint)<br />
or at the hip (torque transmission <strong>with</strong> cable and pulleys).<br />
2.4.2 Results of the optimization<br />
Right at the beginning of the optimization, it turned out that the final velocity<br />
of the hip is slightly higher (3.4 m/s compared to 3.1 m/s) when the mass at<br />
the hip is larger, <strong>with</strong>out having included the velocity losses shown in eq. 2.29<br />
for a bigger mass at the knee. Considering also these losses, it can be followed<br />
that a small mass at the knee is favorably. This led to the final decision on the<br />
design that the cable design <strong>with</strong> pulleys is superior to the motor knee design,<br />
and that this design would be constructed.<br />
Evident became also that for a 5 kg leg, the motors 5 and 6 (RE 25 and RE<br />
26) were too small for the admissible power supply. They pulled only 9.2 A,<br />
and 7.6 A, respectively, when they are at zero speed.<br />
Figure 13: Comparison of motor optimization. The circles indicate the results<br />
<strong>with</strong> the voltage U = 36V<br />
Figure 13 compares the optimization results of the other four motors, when<br />
the mass of the hip is 4 kg and the mass at the knee is 1 kg. The plot to<br />
the upper left shows the velocity of the hip, when the motors are charged <strong>with</strong><br />
the nominal voltage from the catalogue and when they are supplied <strong>with</strong> the
2.4 Optimization 17<br />
Motor 1 2 3 4<br />
h(Unom) [cm] 39 31 33 36<br />
h(U = 36V ) [cm] 54 35 40 36<br />
Table 4: Optimized jump height<br />
maximal voltage of 36 V, which is short time possible for any motor. Motor 4<br />
has its nominal voltage at 36 V. In tab. 4, the end velocity is calculated to the<br />
maximal height of the jump.<br />
Motor 1 (RE 40) reaches the highest velocity value. It speaks against it that<br />
it has a very large mass. The high power consumption comes from the fact that<br />
motor 1 has a large armature resistance, and always runs at the current limit<br />
of 10A when supplied <strong>with</strong> 36 V, indicating that this motor clearly is oversized.<br />
Thus, motor 1 is not considered.<br />
It is remarkable that except for the spring constant k the parameter values<br />
are relatively constant for motor 2, 3, and 4, and the decision which motor to<br />
take is not evident.<br />
Because motor 3, compared to motor 4, jumps slightly higher, the torque<br />
is a bit smaller, and a larger gearbox that can hold 22.5 Nm can be combined<br />
<strong>with</strong> it, which leaves open options for the construction, the decision falls to the<br />
RE 35, outweighing the disadvantage of the larger mass.<br />
Figure 14: Simulation of an optimal launch for a jump<br />
Figure 14 shows the plots of the optimized jump for the DC Motor RE 35.
2.4 Optimization 18<br />
The upper left plot demonstrates the initial compression of the spring leading<br />
to a down movement of the hip, followed by the leg extension. The velocity<br />
curve demonstrates well the acceleration until the moment of take-off.<br />
In the lower left, the absolute value of the force to the ground shows a<br />
maximum of 140 N.<br />
The plot of the current is very characteristic. First it is running at the limit<br />
until the spring is engaged followed by a sharp drop and then a smoothly rising<br />
curve, as the torque rises.<br />
Figure 15: Force vectors of the leg to the ground<br />
The change of force direction before take-off is illustrated in fig. 15. It<br />
is important to see that at take-off, there is a force left into the positive x-<br />
direction, because the shank rotates around the knee joint, and the elasticity is<br />
not yet fully relaxed.<br />
In simulation, k is investigated. When leaving all the parameters the same<br />
and only varying k, the velocity at take-off barely changes, if the variation lies<br />
<strong>with</strong>in ±1.5 Nm/rad.<br />
2.4.3 Selection<br />
The highest jump for 5 kg body mass, was achieved <strong>with</strong> the Maxon motor<br />
RE 35 (30 V) (Section 2.4.2). The matching gearbox holding the maximum<br />
torque is the planetary gearbox GP 42 C. The optimized transmission ratio of
2.4 Optimization 19<br />
1:75 can be achieved <strong>with</strong> a gearbox of that ratio or <strong>with</strong> a gearbox of 1:26 and<br />
additional a transmission ratio <strong>with</strong> pulleys of 1:3.<br />
The maximal torque for the gearbox is 22.5 Nm. This torque defines the<br />
maximal loading during a jump. All mechanical components will be designed<br />
to <strong>with</strong>stand it.<br />
The optimal spring constant lies in the range of 9-12 Nm/rad. Better con-<br />
trollability of a harder spring tipped the scale towards an elasticity of 12 Nm/rad.
3 Mechanical Design 20<br />
3 Mechanical Design<br />
3.1 Introduction<br />
In this Chapter a constructive solution for the robotic leg is introduced. Me-<br />
chanical stability must be adequate to the values found in section 2.4, where the<br />
maximum force on the ground was 140 N and the maximum motor torque was<br />
22.5 Nm. The mechanical construction should be light-weight, and integrate<br />
the motor RE 35 <strong>with</strong> a transmission of around 1:75, that can be realized <strong>with</strong><br />
the gearbox GP 42 C, and an elasticity of 12 Nm/rad.<br />
3.2 Concepts<br />
The key functionalities and partial functionalities are given in the morphological<br />
box in appendix A.2.<br />
The most important decisions in the selection process are explained in the<br />
following sections. The major issues are how to position the motors and what<br />
elastic element to include between shank and knee motor.<br />
3.3 Spring selection<br />
From simulation (Sec. 2.4), we obtained the rotational spring constant k =<br />
12Nm/rad creating a maximum torque of Tmax = 22.5Nm. The maximum<br />
rotation angle is thus φmax = Tmax<br />
k<br />
= 107◦ .<br />
The simplest solution would be to have a torsional spring beside the knee<br />
pulley. But torsional springs that can hold 22.5 Nm and twist 107 ◦ at a rea-<br />
sonable weight could not be found.<br />
Other solutions use a linear extension or compression spring transforming<br />
the rotational into a linear elongation that can easily be done <strong>with</strong> a pulley.<br />
The following equations lead to the depencency between the torsional spring<br />
constant k, the spring constant c for a linear elasticity, and the pulley radius r.<br />
M = k · φ (3.1)<br />
F = c · x (3.2)<br />
x = r · φ (3.3)<br />
k = c · r 2<br />
(3.4)<br />
Equation 3.4 means that many linear springs <strong>with</strong> different spring con-<br />
stants can be drawn into consideration by simply changing the pulley radius.<br />
Of course, this will also change the maximum force in the cable and the neces-
3.4 Gear selection 21<br />
sary spring deformation. A smaller radius always means higher cable tension.<br />
Table 5 shows three possible elasticities from Gutekunst Federn [23].<br />
Spring 1 2 3<br />
Mmax [Nm] 22.5 22.5 22.5<br />
k [Nm/rad] 12 12 12<br />
r [mm] 26 22.9 15.2<br />
c [N/mm] 17.6 22.9 51.5<br />
Fmax [N] 865 983 1480<br />
∆l [mm] 49 43 29<br />
l0 [mm] 150 110 76<br />
d [mm] 20 32 33<br />
m [g] 130 165 148<br />
Table 5: Comparison of three linear compression springs: radius of the pulley<br />
r, spring constants k and c, pulley radius r, maximum force Fmax, elongation<br />
∆l, relaxed spring length l0, spring diameter d, mass m<br />
Spring 2 (App. A.3) was selected. It is shorter than spring 1, and the force<br />
is not as large as in spring 3. The disadvantage is the mass, but compared to<br />
other springs on the market it is still rather light weight.<br />
Extension springs are drawn into consideration, too, but as they extend<br />
under force, they are too long to fit into the shank.<br />
3.4 Gear selection<br />
Considered for motor gearboxes were Maxon planetary gears, harmonic drives,<br />
and cable pulley constructions.<br />
Harmonic drives have very high transmission ratios and no backlash but<br />
were too heavy. So that planetary gears matching the Maxon motors are chosen.<br />
Because pulleys were used anyway to apply torque to the knee provided from<br />
a motor at the hip, a ratio of 1 to 3 of the pulley radi was added between<br />
the hip pulley and the knee pulley which allowed the use of a smaller gearbox<br />
(transmission ratio 1:26), therefore saving weight and efficiency.<br />
The hip joint has two degrees of freedom: swinging forward/backward and<br />
swinging to the side. A construction we call ’differential drive’ was designed. In<br />
fig. 16 the functioning principle is shown. When the first hip motor T1 and the<br />
second hip motor T2 turn in the same direction the leg moves sideways (torque<br />
Thx). During a counter rotation, the torque Thz is built up that results in a<br />
forward/backward swing of the leg.<br />
The advantages over a construction that separates the two degrees of free-<br />
dom lies mainly in the possibility to have two smaller motors working together
3.4 Gear selection 22<br />
(a) Concept (b) Picture<br />
Figure 16: Differential drive<br />
for a movement. A leg swings a lot forward and backwards. If the two degrees<br />
of freedom were decoupled, one motor would stay passive and must still be<br />
carried around. Two motors working together must provide each only half of<br />
the necessary torque reducing the size of motor and gearbox. Another advan-<br />
tage is that both motors can be fixed to the body <strong>with</strong>out losing one degree<br />
of freedom. The added complexity to the construction is the major drawback<br />
of the differential drive assembly. Other disadvantages are the weight, size and<br />
the backlash of the bevel gear.<br />
Figure 17: Load estimation hip: Offset torque F · lh because the leg is not<br />
positioned on the axis of the differential drive<br />
To check which option saves more weight a loading scenario is calculated.<br />
Figure 17 shows the force F from the leg that accelerates the hip upwards.<br />
Because this force is not acting on the axis of the differential drive the two hip<br />
motors have to hold against the resulting torque Thx given by<br />
Thx = lh · F = 10Nm (3.5)<br />
when lh is 5 cm and the maximum force is estimated to be 200 N, which includes<br />
a safety of 1.5 to the force that resulted from the jump analysis in section 2.4.2.<br />
If two motors provide this torque, each has to provide 7.5 Nm. This has
3.5 Hip 23<br />
Figure 18: Hip design<br />
to be held also by the bevel gears of the differential drive. Table 3 shows that<br />
from Maxon the gearbox GP 32 C has a maximum torque of 7.5 Nm <strong>with</strong> the<br />
suitable motor RE 25. Nozag, a bevel gear manufacturer, offers spiral bevel<br />
gears that can transmit 10 Nm, seen from the small bevel gear (App. A.3).<br />
The teeth of the bevel gear are the critical point, so that two gears attaching<br />
at different teeth do not influence the maximal torque. Together they can<br />
hold Thx = 15Nm. The exceeding torque can therefore be used for dynamic<br />
movements like actuating the thigh during a jump and not only to hold the hip<br />
in place statically.<br />
With the transmission of 1:2.5 inside the gearbox the maximum torque in<br />
the forward and backward swing becomes 37.5 Nm, which is sufficient for any<br />
loading scenario for this leg.<br />
This option <strong>with</strong> two motors and three bevel gears weighs 1.11 kg (650 g+460 g).<br />
If the two motors are decoupled, the gearboxes must provide alone 15 Nm<br />
and 37.5 Nm. From the Maxon catalogue, the GP 42 C and the GP 52 C <strong>with</strong><br />
a suitable motor achieve the same strength. These two larger motors <strong>with</strong> gears<br />
weigh together 2.05 kg (1.25 kg+700 kg).<br />
Due to the much lower mass and the advantages described before, the deci-<br />
sion falls towards the differential drive.<br />
Having the answers to these three major decisions - what spring, what gears<br />
and what motors to use -, the task was now to find a suitable construction<br />
embedding these components.
3.5 Hip 24<br />
3.5 Hip<br />
Figure 19: Hip detail 1<br />
The hip (Fig. 18) combines all three degrees of freedom -two of the hip and one<br />
of the knee- in a very compact design.<br />
The hole of the bevel gear was enlarged and the housing of the knee motor<br />
is attached directly to the large bevel gear (Fig 19). Between motor shaft<br />
and the shaft extension <strong>with</strong> an integrated pulley, a coupling was inserted that<br />
compensates for small angle or displacement deviations.<br />
Enclosing the shaft extension a hollow shaft (Fig. 20) was designed. It is<br />
also attached directly to the large bevel gear and supports the shaft extension<br />
of the motor. This support is needed because the radial force on the pulley<br />
(shaft extension) is very large (983 N, Sec. 3.3). The hollow shaft is connected<br />
rigidly to the side plates of the thigh. Therefore, the movement of the upper<br />
leg is coupled directly to the large bevel gear. The cables enter through two<br />
holes into the hollow shaft, where they are wound up around the pulley.<br />
The small bevel gears are actuated by by two smaller motors (Maxon RE 25)<br />
that are connected via torsion springs to the small bevel gears (Fig. 21).<br />
An aluminium profile from Kanya[25] holds the two small motors and the<br />
shafts that support the differential box. In the test setup (Section 5.2) the<br />
profile is constrained to only up and down movement.<br />
3.6 Thigh<br />
The largest challenge of the thigh (Fig. 22(a)) design was the connection to<br />
the hip and to the shank. The shank has a given width because of the spring.<br />
When folding the leg the shank and the cable that is tightened from hip to knee<br />
must neither touch the bevel gear at the hip, nor the tube of the shank at the<br />
knee.
3.7 Knee 25<br />
(a) Cutview hip<br />
(b) Housing of the differential drive<br />
Figure 20: Hip detail 2<br />
The two side plates of the thigh are 44 mm apart, and the cable is guided<br />
around a shaft so that the shank (width 42 mm) can fold close to the thigh.<br />
The braces holding the shaft can be used at the same time to tighten the cable.<br />
Figure 22(b) shows the guidance of the cable around the pulleys and the three<br />
small shafts that guide the cable. The mid part of the thigh has a U-shape<br />
giving the thigh torsional and bending stiffness.<br />
3.7 Knee<br />
The knee axle connects the thigh and the knee (Fig. 23(a)). It is blocked from<br />
rotating inside the shank by a pin through the shaft, so that a potentiometer<br />
can be placed on the shaft measuring the relative rotation against the thigh<br />
(Fig. 23(b)).<br />
The pulley is supported by two ball bearings on the knee shaft. Its diameter<br />
is 56 mm as described in section 3.3 (spring 2). The cable is wound as shown<br />
in fig. 24. The two endings from the shank are secured to the pulley after a<br />
turning radius of 240 ◦ . The two cable endings leading from the hip shaft to the
3.8 Shank 26<br />
Figure 21: Actuation of the differential drive<br />
knee are wound 1.5 times around the pulley (if leg is stretched) and secured<br />
there to the pulley.<br />
At the fastening point, the cable leads through a hole into the the pulley<br />
and a knot on the other side of the hole prevents the cable from gliding back.<br />
3.8 Shank<br />
Inside the shank, the compression spring is supported by two plates. The<br />
bottom plate close to the foot can slide up and down and has two holes where<br />
the cable is attached. To reduce the impacts on the bottom plate, when the<br />
compression spring is rapidly relaxed and compressed, a small rubber damper<br />
is located at the fastening point of the cable.<br />
The foot is a simple half sphere that can smoothly roll on the floor.<br />
3.9 Cable<br />
Dyneema was selected for cable. The mechanical properties outperform steel<br />
or other composites (Fig. 26)<br />
3.10 Sensors<br />
Encoders measure the angular position of the motor. The RE 25 use an optical<br />
encoder and the RE 35 a magnetoresistive one, which is smaller, so that the<br />
bush supporting the motor can be slid over the encoder.<br />
Three potentiometer measure the rotation angle of the shank relative to the<br />
thigh, and the rotation of the smaller bevel gears relative to the hip.<br />
With the difference between the encoder and the potentiometer signals, the<br />
spring deflection can be measured.
3.10 Sensors 27<br />
(a) Thigh<br />
(b) Cable guidance<br />
Figure 22: Thigh
3.10 Sensors 28<br />
(a)<br />
(b)<br />
Figure 23: Knee
3.10 Sensors 29<br />
Figure 24: Cable winding around the knee pulley<br />
(a) Shank (b) Cut view<br />
Figure 25: Shank<br />
Figure 26: Dyneema, source: www.swiss-composite.ch/pdf/t-dyneema.pdf
4 Control 30<br />
4 Control<br />
4.1 Introduction<br />
In Chapter 2 a simple one-motor model has been introduced to estimate forces,<br />
torques and to select motors, gearboxes and elastic elements. This model is<br />
extended in this Chapter. The geometric and mass properties of the actual leg<br />
developed in Chapter 3 is included, and the leg is actuated by all three motors.<br />
An optimal feedforward control algorithm is proposed to realize a high jump,<br />
and the influence of a combined knee and hip actuation is analyzed.<br />
4.2 Improved model<br />
To control the motors, Maxon EPOS controllers were used. These controllers<br />
allow direct control of position, velocity, acceleration and deceleration. The<br />
lower level control of current or voltage was done by the EPOS.<br />
This means that the input variables of the model are the motor angles φ1, φ2<br />
and φ3 on one end of the springs. This simplifies the complexity of the model.<br />
The output variables remain the positions and the velocities of the bodies.<br />
The leg is built of three major parts.<br />
• Body 1 includes the aluminium profile <strong>with</strong> the two hip motors and the<br />
smaller bevel gears. All these parts will be summarized as hip.<br />
• Body 2 consists of the large bevel gear <strong>with</strong> the shaft and the knee motor<br />
fixed to it, the parts of the thigh, and the pulley at the knee. This<br />
assembly is referred to has thigh.<br />
• The knee elasticity, the knee, the tube of the shank and the foot are<br />
assembled to the shank forming body 3.<br />
To define the initial state, the kinematic equations were derived to relate the<br />
motor angles to the coordinates of the ground contact point and of the hip.<br />
The geometrical properties of the parts are described in fig. 27. The inertial<br />
reference frame is positioned vertically underneath the center of gravity of the<br />
hip.
4.2 Improved model 31<br />
<strong>with</strong><br />
Figure 27: Geometry and coordinate system of the leg<br />
x0 = 0 (4.1)<br />
y0(φ1, φ2, φ3) = − cos φhx · (−l sin φhz − l sin φk) + lh sin φhx<br />
(4.2)<br />
z0 = 0 (4.3)<br />
xg(φ1, φ2, φ3) = l cos φhz − l cos φk<br />
(4.4)<br />
yg = 0 (4.5)<br />
zg(φ1, φ2, φ3) = sin φhx · (−l sin φhz − l sin φk) + lh cos φhx<br />
φhz = φ1 − φ2<br />
2ndiff<br />
φhx = φ1 + φ2<br />
2<br />
φk = φ3 + φhz<br />
(4.6)<br />
(4.7)<br />
(4.8)<br />
(4.9)<br />
φhz and φhx describe the rotation around the x- and z-axis of the hip and φk<br />
the rotation of knee in respect to the inertial reference frame. The transmission<br />
of the bevel gears in the differential drive is ndiff .<br />
The mass properties of the three bodies are summarized in tab. 6. The hip<br />
itself has only one translational degree of freedom, thus the rotation is blocked.<br />
The inertias Jh1 and Jh2 describe only the inertia of the motor shaft and the<br />
bevel gear.<br />
The analytical derivation of the dynamic model using for example the La-
4.2 Improved model 32<br />
Mass [kg] Inertia [10 − 3 kg m 2 ]<br />
mh = 2.2 Jh1 = Jh2 =<br />
mt = 1.7 Jt =<br />
ms = 0.4 Js =<br />
⎛<br />
1.8 0 0<br />
⎞<br />
⎝ 0 37.3 0 ⎠<br />
⎛<br />
0.2<br />
0 0 37.3<br />
⎞<br />
−0.2 −3.6<br />
⎝ −0.2 13.2 0.1 ⎠<br />
−3.6<br />
⎛<br />
0.1<br />
0.1<br />
0<br />
7.8<br />
⎞<br />
0<br />
⎝ 0 1.2 0 ⎠<br />
0 0 1.2<br />
Table 6: Mass properties of the leg identified from the 3D-CAD model<br />
grange method becomes too complex so that the numerical Matlab toolbox for<br />
mechanical systems was used again. SimMechanics evaluates these geometric<br />
and mass properties (Fig. 28).<br />
Figure 28: SimMechanics model of the leg<br />
The three series elastic actuators, each containing a motor, a transmission<br />
and a spring, actuate the hip around the x- and the z-axis, and the knee around<br />
the z-axis.<br />
With the torques of the motors T1, T2, and T3, the actuation around the<br />
three axes can be derived as<br />
Thz = T1 − T2<br />
ndiff<br />
Thx = T1 + T2<br />
Tk = T3<br />
(4.10)<br />
(4.11)<br />
(4.12)
4.3 Optimized feedforward control 33<br />
4.3 Optimized feedforward control<br />
4.3.1 Standing<br />
Standing appears to be an easy task. But in fact, this leg configuration is always<br />
at a instable position. This can be seen in the following function<br />
Tk(φ) = mglcos(φ) (4.13)<br />
Tk is the required torque at the knee depending on the angle φ. m is the mass,<br />
g the gravity constant and l the length of the thigh. Tk is defined in eq. 4.13,<br />
but is simplified in this respect that the mass is distributed in all three bodies<br />
and not only in the hip. The mass m is therefore substituted by the mass m ∗ ,<br />
which is determined in simulation. m ∗ is slightly smaller than m but is constant<br />
and independent of φ.<br />
If the torque from the motor is too low, the leg moves downwards (φ de-<br />
creases), where an even higher torque would be necessary, so the leg drops to<br />
the floor. The inverse happens if the torque is slightly too high. The leg ex-<br />
pands (φ grows) to an angle where the torque should be even lower. The result<br />
is that the leg goes into full stretching. This means that the motor torque to<br />
stand still has always to be controlled in a loop. The Maxon EPOS controller<br />
does this if the angle and not the torque is controlled.<br />
It is important to know what the initial motor position needs to be for a<br />
required initial leg angle.<br />
φm =<br />
4.3.2 Optimization algorithm<br />
�<br />
φ + Tk<br />
�<br />
· γ (4.14)<br />
ks<br />
The goal to perform the highest possible jump is equivalent to maximizing the<br />
final vertical velocity before take-off of the leg. For optimizing, constraints must<br />
be considered. In this model there are three constraints that are explained next.<br />
• Motor properties<br />
• Foot is not allowed to slip<br />
• Foot can only push towards the ground<br />
Motors cannot turn at all speeds, and the maximum torque and the accel-<br />
eration is limited. Angular velocity ω and maximum torque Tmax obey a linear<br />
relationship. The back emf as described in section 2.3.1 increases linearly <strong>with</strong>
4.3 Optimized feedforward control 34<br />
ω, lowering Tmax.<br />
Tmax = κa · Ia,max − κi · ω (4.15)<br />
The maximal torque is limited by the mechanical construction to 22.5 Nm.<br />
Figure 29: Power constraint of the motor<br />
The transmission ratio γ of the knee actuator of 1 : 78, thus the maximum<br />
torque needed from the motor is 0.29 Nm, which can be supplied until the knee<br />
pulley turns at n = 5634<br />
78<br />
= 72.2 rpm or 7.6 rad/s (Fig. 29). If the motor turns<br />
faster the maximum torque is reduced. The motor constraint becomes<br />
<strong>with</strong><br />
Tmax =<br />
�<br />
Tsim ≤ Tmax<br />
22.5Nm, if n ≤ 7.6 rad<br />
s<br />
T0 − κi · ω if n > 7.6 rad<br />
s<br />
The motor velocity is ω, and the stalling torque is T0.<br />
�<br />
(4.16)<br />
(4.17)<br />
The dynamics of the inductance inside the motor are not accounted for in<br />
this model. It must be checked later in experiments, if this is relevant for the<br />
leg application (Section 5.3.2).<br />
The leg dynamics and the motor torques do not only induce a vertical<br />
force (y-axis) on the ground but also parallel forces. The second constraint is<br />
therefore that the friction holding the foot in place must be always larger than<br />
the horizontal forces. If µ is the friction coefficient the constraint can be written<br />
as<br />
<strong>with</strong><br />
FF ≤ µ |Fy| (4.18)<br />
FF = � F 2 x + F 2 z<br />
(4.19)
4.3 Optimized feedforward control 35<br />
The last constraint is the stop criterion of the simulation. If the vertical<br />
force on the ground becomes positive - foot wants to pull on the ground - which<br />
is not possible. The simulation stops, and tend and vy(tend) are determined.<br />
The following cost functional J for the optimization showed good results<br />
J = −vy(tend) − v 3 � tend<br />
y(tend) + cF<br />
0<br />
j�<br />
cT jdt (4.20)<br />
j being the number of motors used. cT and cF follow from the constraints and<br />
are defined by<br />
�<br />
(Tsim − Tmax)<br />
cT =<br />
k �<br />
, if Tsim > Tmax<br />
0, if Tsim ≤ Tmax<br />
�<br />
(FF − µ |Fy|)<br />
cF =<br />
k �<br />
, if µ |Fy| ≤ FF<br />
0, if µ |Fy| ≥ FF<br />
1<br />
(4.21)<br />
(4.22)<br />
The power factor k is chosen to be 4 so that large exceeding of the maximal<br />
torque increases cT by a power of four and not only linear. With this the<br />
transition is continuous at the boundary of the constrained space and increases<br />
fast when a constrained space is left by more than one unit.<br />
The input variables to the model are the motor angles φ1(t), φ2(t) and φ3(t).<br />
We know from section 2.4.2 that in the optimal trajectory the hip moves down<br />
and accelerates then upwards. This trajectory means for the knee motor to<br />
turn first in one direction and then to accelerate in the other. This behaviour<br />
is approximated by a third order polynomial.<br />
φ3(t) = a03 + a13t + a23t 2 + a33t 3 + φ03<br />
(4.23)<br />
φ03 is the initial motor angle (Eq. 4.14) so that <strong>with</strong>out any other input the leg<br />
would be standing still.<br />
A negative a03 means that the torque in the spring is reduced right at<br />
the beginning, when the motor is switched on, which provokes the initial and<br />
characteristic downwards movement of the hip before accelerating upwards.<br />
For the hip motors (j = 1, 2) the input trajectory is defined by a polynomial<br />
as well.<br />
φi(t) = a0i + a1it + a2it 2 + a3it 3<br />
(4.24)<br />
The twelve coefficients aij and the initial angle of the leg φinit are the<br />
parameter to be optimized inside the algorithm.
4.3 Optimized feedforward control 36<br />
All three motors should work in a way together that they give the most<br />
thrust upwards and at the same time do not work against each other. In the<br />
following subsection the influence of the hip motors is investigated.<br />
4.3.3 Results<br />
At first, the two hip torques were set to zero, and only the torque to the knee<br />
was optimized, starting <strong>with</strong> three optimization parameters: φinit, a03 and a13.<br />
Then the number of parameters were increased by a23 and a33. The foot position<br />
was set exactly underneath the center of gravity of the hip. This tilts the hip<br />
when the knee bends. Another option would be to set the leg into a vertical<br />
position, which would set the foot 4.5 cm besides the other point. This distance<br />
is introduced by the hollow shaft at the hip. The optimization results differed<br />
by about 1 cm jump height or 3 N force, which can be neglected.<br />
In the simulation, a33 has almost no effect on the maximal jump height. For<br />
this reason, a33 was set to zero to keep the number of optimized parameters low.<br />
Figure 30 shows the the angular position (the optimized curve), the velocity<br />
trajectory, and the force trajectories of the optimized jump using only the knee<br />
motor. The maximum velocity of the hip becomes 3.23 m/s, which corresponds<br />
to a jump height of 42 cm. The calculation of the jump height takes into account<br />
that the thigh and the shank have other velocities (Eq. 2.28). The maximal<br />
force towards the ground is 164 N.<br />
The next optimization step was to include the hip motors (i = 1,2). The op-<br />
timized parameters from the knee motor were kept constant, while the number<br />
of parameters (a0i ... a3i) for the hip motor were increased.<br />
From the simulation it could be followed that a0i and a1i have no effect on<br />
the outcome. The output of the optimization set them close to zero, and the<br />
jump height decreased.<br />
It is different for the parameters a2i and a3i, which do influence the jump<br />
height positively. After optimizing the four knee motor parameter and the four<br />
hip motor parameters all together (φinit, a03, a13, a23, a2i, a3i), the optimized<br />
curve in fig. 31 resulted.<br />
The velocity at take-off is 3.39 m/s and the calculated jump height is 46 cm.<br />
It follows that in comparison to the case where only the knee was actuated, the<br />
hip motors had only a slight influence on the jump(4 cm higher). One reason<br />
might be that at some point they supported the jump action whereas at other<br />
they reduced the force. This because the each hip motor trajectory was defined<br />
by only two parameters.
4.4 Implementation 37<br />
(a) Angular position of the knee motor (b) Position and velocity trajectories of<br />
the hip<br />
(c) Force trajectory on the ground (d) x-, y-, z- components of force<br />
Figure 30: Optimized trajectories if only the knee motor applies a torque<br />
4.4 Implementation<br />
To find the optimal trajectories in simulation is only one part to the control<br />
of the motors. It is also important that the communication <strong>with</strong> the hardware<br />
works smoothly. The critical points here is the frequency of the data exchange.<br />
We have three options<br />
• Maxon software <strong>with</strong> RS 232-CAN cable connection<br />
• Matlab <strong>with</strong> RS 232-CAN cable connection<br />
• Matlab <strong>with</strong> CAN-CAN cable connection<br />
The software from Maxon is very reliable and the user interface is very<br />
easy and intuitive to use. The ASL developed a Matlab code using the same<br />
functions as the Maxon software. The data exchange rate is not increased<br />
compared to the Maxon software when using the RS 232 bus, but trajectories<br />
can be implemented for forward control. The CAN-CAN connection has a<br />
remarkably higher data exchange rate, and of course the trajectories can be
4.4 Implementation 38<br />
(a) Angular position of the motors (b) Position and velocity trajectories of<br />
the hip<br />
(c) Force trajectory on the ground (d) x-, y-, z- components of force<br />
Figure 31: Optimized trajectories if all motors apply torque<br />
implemented just like in the second option. The disadvantage was, that the<br />
software for the CAN-CAN connection was not very stable, and still under<br />
developement.
5 Assembly and Test 39<br />
5 Assembly and Test<br />
5.1 Assembly<br />
The leg could be assembled completely, though a few parts need to be modified<br />
to enable a smooth assembly of the leg.<br />
The circlip on one side of the knee cannot be mounted because the support<br />
of the bearing to the thigh is protruding. The border securing the ball bear-<br />
ings should be on the other side of the insert, which is supporting the bearing<br />
(Fig. 32).<br />
Figure 32: Circlip cannot be mounted to the knee shaft<br />
The pin in the shaft of the knee to block the shaft from rotating does not<br />
fit by 0.5 mm. The smaller pin that was used introduced several degrees of<br />
backlash.<br />
The torsion spring does not fit well into the groove of the coupling because<br />
the radius of the spring is larger then expected. This coupling needs a whole<br />
redesign (Fig. 33).<br />
The support of the shaft extension of the hip motors block the leg from ro-<br />
tating around the x-axis (rotation sideways), when the leg is completely folded.<br />
The corners of this support block can be chamfered (Fig. 34) to prevent this.<br />
The lower support of the compression spring slides up and down inside the<br />
tube of the shank. The support and the tube are made of aluminum which is<br />
not preferable, because this introduces remarkable friction losses. Possibly the<br />
support can be made of a composite (Fig. 25).<br />
The cable ripped initially at the edge where the cable is deflected to wind<br />
up around the pulley. Thick double-sided adhesive tape on the pulley increased<br />
the friction on the pulley and decreasing the strain over the edge (Fig. 35).
5.1 Assembly 40<br />
Figure 33: Torsion spring does not fit into the groove<br />
Figure 34: Corner should be chamfered
5.1 Assembly 41<br />
Figure 35: Adhesive tape increases the friction of the pulley to avoid damage<br />
to the cable
5.2 Test setup 42<br />
5.2 Test setup<br />
To test the leg to stand, to move up and down, and to jump, the aluminium<br />
profile of the hip was attached to another profile of 1.5 m length leading to<br />
the wall. It was secured there <strong>with</strong> hinge joints so that it could move on a<br />
circular arc up and down leaving only one degree of freedom unconstrained.<br />
The consequence of this defined path of the hip requires the hip to be able to<br />
freely rotate around the x-axis. If this rotation were blocked the torsion springs<br />
of the hip motors would be wound up creating a force towards the side of the<br />
foot, which in the end brakes the acceleration.<br />
Figure 36: Test setup: mounting<br />
To prevent slipping of the foot, it was covered <strong>with</strong> a sticky coating, and<br />
the ground was covered <strong>with</strong> a rubber mat (Fig. 36). This absorbeded also<br />
potential side forces created by the hip motor. The controllers were set beside<br />
the leg on the ground, and not mounted to the leg.<br />
The whole leg including the profile leading to the wall but <strong>with</strong>out the EPOS<br />
controller weighs 4.3 kg. With the controller, the leg and the aluminium profile<br />
have a mass of 5 kg.
5.3 Results 43<br />
5.2.1 Motor control<br />
The two hip motors were connected to the EPOS 24/5, and the knee motor to<br />
the EPOS 70/10. The connection from the computer to the EPOS was either<br />
using the RS 232 bus or the CAN bus (Section 4.4).<br />
It resulted that the EPOS 70/10 cannot be combined in series <strong>with</strong> the<br />
EPOS 24/5, so that for the tests either three EPOS 24/5 can be used when all<br />
motors are mounted, or one EPOS 70/10 if only the knee motor is controlled<br />
and both hip motors are dismounted.<br />
To program the EPOS, like setting the controller gains, or defining the<br />
maximum acceleration the software provided from Maxon proved to be very<br />
useful. To follow a position trajectory the control using Matlab is necessary.<br />
It was first tested, if the motor is able to follow the optimal trajectory<br />
calculated in section 4.3.3. The connection used the RS 232 bus. First, the<br />
trajectory was stretched, meaning that the same trajectory points had to be<br />
passed, but in larger time intervals. At lower speed the desired trajectory was<br />
perfectly matched by the motor. But at higher speed the motor was too slow<br />
(Fig. 37).<br />
Two possibilities exist to explain this. Either the motor is not able to<br />
accelerate so quickly, or the data exchange rate is too low.<br />
To increase the data exchange rate, a connection via the CAN bus could be<br />
established.<br />
5.3 Results<br />
5.3.1 Preliminary tests <strong>with</strong> the leg<br />
The leg was positioned <strong>with</strong> a 90 ◦ angle at the knee, and the foot lying under-<br />
neath the hip (Fig. 36). Slowly, the hip was moved up and down.<br />
It became evident that the cable lengthens by several centimeters because<br />
it is braided. It did not seem to be elastic, though. This makes it necessary to<br />
retension the cable several time after it is tightened the first time.<br />
The cable is vulnerable at the points where it is guided over an edge, namely<br />
at the position where it is guided through the holes of the pulley (Fig. 35).<br />
Already at slow movement, the cable ruptures. This makes it necessary to<br />
increase the friction on the pulley, so that the larger part of the force in the<br />
cable is absorbed by friction.
5.3 Results 44<br />
5.3.2 <strong>Maximum</strong> jump<br />
As a first approach to realize a jump a start at the lowest position was chosen,<br />
and then the leg was stretched <strong>with</strong> maximal power. The hip motors were not<br />
used and therefore removed from the leg. The knee motor was controlled <strong>with</strong><br />
the EPOS 70/10.<br />
For this setup the provided software from Maxon was sufficient. The end<br />
position was an almost stretched leg (knee angle ≈170 ◦ ). And by increasing<br />
the parameters of the motor control (current, voltage, velocity and acceleration<br />
limitation) maximum power is approached. A final end position at less than<br />
170 ◦ slowed down the extending leg too soon so that a jump was not possible.<br />
At a voltage of 31 V and a current limit of 5.5 A, the leg lifted off the floor<br />
(Fig. 38). After take-off, the thigh kept on rotating further around the hip.<br />
The shank followed the thigh. It must be noticed that for this jump, there were<br />
no hip motors mounted to stop this swinging movement.<br />
In Fig. 38(i) two lines indicate the position of the aluminium profile at the<br />
momement of take-off and at maximum altitude. From the picture the jumping<br />
height can be estimated to 5.5 to 6 cm, which corresponds to the distance from<br />
the top of the aluminium profile to the middle on the side of of the motor<br />
support plate.<br />
There are two main reasons why the 42 cm from simulation for one motor<br />
(Sec. 4.3.3) was not reached. The leg was actuated <strong>with</strong> maximally 5.5 A<br />
which is approximately one half of the power that was used in the simulation.<br />
Secondly, the optimized control trajectory was not implemented. Inaccuracies<br />
in the model and frictional losses might have reduced the maximal height, too.
5.3 Results 45<br />
(a) Trajectory control twenty times slower<br />
(b) Trajectory control two times slower<br />
(c) Trajectory control normal speed<br />
Figure 37: Trajectory control
5.3 Results 46<br />
(a) (b) (c)<br />
(d) (e) Lift off (f)<br />
(g) (h) Top (i) Max. height 5.5-6 cm<br />
Figure 38: Picture series of a jump
6 Summary and Future Work 47<br />
6 Summary and Future Work<br />
6.1 Summary<br />
In this thesis, four mechanical designs for a robotic leg were evaluated. The<br />
selected design option had the best motor torque to force ratio. To set most<br />
mass at the hip (like the knee actuation motor) resulted in optimization to be<br />
better then locating the motor at the knee. The torque transmission from the<br />
motor at the hip to the knee has been realized <strong>with</strong> a cable pulley construction.<br />
The leg was constructed <strong>with</strong> a series elastic actuator, to achieve maximum<br />
peak power, to recover energy, and to increase the shock tolerance of the gear<br />
and the motor. For the knee actuation, the spring and the motor <strong>with</strong> a gearbox<br />
was found after optimization. A compression spring has been integrated into<br />
the shank so that <strong>with</strong> the knee pulley a transformation from a linear to a<br />
rotational elasticity could be realized.<br />
A model <strong>with</strong> the geometric and mass properties of the leg has been used to<br />
find and to optimize feedforward control trajectories for the knee and the hip<br />
motors. In simulation, a maximum jump height of 46 cm was reached.<br />
In first tests, the construction showed to be operating reliably. The leg<br />
jumped 5.5 to 6 cm centimeters, starting from a folded leg position, and ac-<br />
celerating <strong>with</strong> maximum available power. The leg did not jump 46 cm, as<br />
predicted in the simulation. Reasons are the limited power supply of 31 V and<br />
5.5 A. Also, the optimized trajectory has not yet been implemented. Another<br />
reason is the inaccuracy of the model, due to losses of the backlash in the trans-<br />
mission, friction of bearings, of the mounting to the wall, and of the spring and<br />
the support plate inside the shank, which are not included in the model.<br />
6.2 Future work<br />
Mechanical improvements to reduce the weight and to alter or redesign the<br />
parts described in section 5.1 must be made.<br />
An Evaluation and comparison of the model and the actual leg will give<br />
further insights to the dynamical behavior of the leg.<br />
The feedforward control algorithm can be implemented, and enhanced by<br />
more sophisticated control input or by feedback control. The developement of<br />
a control strategy after take-off is necessary to have a controlled landing and<br />
to make a series of jumps possible.<br />
Ultimately, this thesis can be extended to other dynamic scenarios of or<br />
more leg.
A Appendix 48<br />
A Appendix<br />
A.1 Derivation of kinematic equations for the pantographic design<br />
The equations for the coordinate xC and yC originate from the calculation of<br />
the intersection of two circles [26].<br />
� � � �<br />
A =<br />
�<br />
xA<br />
yA<br />
�<br />
0<br />
=<br />
0<br />
� �<br />
B<br />
xB(φ)<br />
=<br />
yB(φ)<br />
� �<br />
l sin φ<br />
=<br />
−l cos φ<br />
�<br />
�<br />
D =<br />
xD(φ)<br />
yD(φ)<br />
2d sin φ2<br />
=<br />
−2d cos φ2<br />
�<br />
p =<br />
a = l2 − d 2 + p 2<br />
(xB − xD) 2 + (yB − yD) 2<br />
2p<br />
(A.1)<br />
(A.2)<br />
(A.3)<br />
(A.4)<br />
(A.5)<br />
h = � l2 − a2 (A.6)<br />
� � �<br />
xC(φ) xD +<br />
C =<br />
=<br />
yC(φ)<br />
a<br />
p (xB − xD) + h<br />
p (yD − yB)<br />
yD + a<br />
p (yB − yD) + h<br />
p (xB<br />
�<br />
(A.7)<br />
− xD)<br />
� � �<br />
xE(φ) xC −<br />
E =<br />
=<br />
yE(φ)<br />
|xB−xC|<br />
�<br />
d · (l + d)<br />
(A.8)<br />
· (l + d)<br />
y(φ) = yC −<br />
yC − |yB−yC|<br />
d<br />
l + d<br />
d<br />
(A.9)<br />
· |−l cos φ − yC| (A.10)
A.2 Morphological box 49<br />
A.2 Morphological box
A.2 Morphological box 50
A.2 Morphological box 51
A.2 Morphological box 52
A.3 Data sheets 53<br />
A.3 Data sheets<br />
• Maxon RE 35<br />
• Maxon GP 42 C<br />
• Maxon RE 25<br />
• Maxon GP 32 C<br />
• Gutekunst Federn D-357<br />
• Federntechnik Knoerzer M45R21<br />
• Nozag SS 131435<br />
• Maedler Ausgleichskupplung LA 60293600
A.3 Data sheets 54<br />
RE 35 �35 mm, Graphitbürsten, 90 Watt<br />
M 1:2<br />
Lagerprogramm<br />
Standardprogramm<br />
Sonderprogramm (auf Anfrage)<br />
Bestellnummern<br />
gemäss Massbild 273752 323890 273753 273754 273755 273756 273757 273758 273759 273760 273761 273762 273763<br />
Wellenlänge 15.6 gekürzt auf 4 mm 285785 323891 285786 285787 285788 285789 285790 285791 285792 285793 285794 285795 285796<br />
Motordaten (provisorisch)<br />
Werte bei Nennspannung<br />
1 Nennspannung V 15.0 24.0 30.0 42.0 48.0 48.0 48.0 48.0 48.0 48.0 48.0 48.0 48.0<br />
2 Leerlaufdrehzahl min-1 7070 7670 7220 7530 7270 6650 5960 4740 3810 3140 2570 2100 1620<br />
3 Leerlaufstrom mA 245 168 123 92.7 77.3 68.7 59.7 44.7 34.2 27.1 21.6 17.2 12.9<br />
4 Nenndrehzahl min-1 6270 6910 6420 6770 6490 5860 5150 3920 2970 2280 1710 1220 732<br />
5 Nennmoment (max. Dauerdrehmoment) mNm 73.2 93.3 92.4 97.7 96.5 98.2 98.8 102 105 105 105 104 104<br />
6 Nennstrom (max. Dauerbelastungsstrom) A 4.00 3.36 2.50 1.95 1.63 1.51 1.36 1.12 0.915 0.752 0.621 0.503 0.391<br />
7 Anhaltemoment mNm 874 1160 949 1070 967 878 766 613 493 394 320 253 194<br />
8 Anlaufstrom A 45.0 39.7 24.4 20.3 15.5 12.9 10.1 6.43 4.16 2.74 1.83 1.18 0.704<br />
9 Max. Wirkungsgrad<br />
Kenndaten<br />
% 81 84 84 86 85 85 84 83 82 80 79 77 74<br />
10 Anschlusswiderstand � 0.334 0.605 1.23 2.07 3.09 3.72 4.75 7.46 11.5 17.5 26.2 40.5 68.2<br />
11 Anschlussinduktivität mH 0.085 0.191 0.340 0.620 0.870 1.04 1.29 2.04 3.16 4.65 6.89 10.3 17.1<br />
12 Drehmomentkonstante mNm A-1 19.4 29.2 38.9 52.5 62.2 68 75.8 95.2 119 144 175 214 276<br />
13 Drehzahlkonstante min-1 V-1 491 328 246 182 154 140 126 100 80.5 66.4 54.6 44.7 34.6<br />
14 Kennliniensteigung min-1 mNm-1 8.43 6.79 7.76 7.16 7.62 7.67 7.89 7.85 7.84 8.08 8.19 8.46 8.55<br />
15 Mechanische Anlaufzeitkonstante ms 5.97 5.60 5.50 5.40 5.38 5.38 5.39 5.38 5.37 5.38 5.39 5.39 5.41<br />
16 Rotorträgheitsmoment gcm2 67.6 78.7 67.6 72.0 67.4 67.0 65.2 65.4 65.5 63.6 62.8 60.8 60.4<br />
Spezifikationen<br />
Thermische Daten<br />
17 Therm. Widerstand Gehäuse-Luft 6.2 KW -1<br />
18 Therm. Widerstand Wicklung-Gehäuse 2.0 KW -1<br />
19 Therm. Zeitkonstante der Wicklung 30 s<br />
20 Therm. Zeitkonstante des Motors 1050 s<br />
21 Umgebungstemperatur -20 ... +100°C<br />
22 Max. Wicklungstemperatur +155°C<br />
Mechanische Daten (Kugellager)<br />
23 Grenzdrehzahl 12000 min -1<br />
24 Axialspiel 0.05 - 0.15 mm<br />
25 Radialspiel 0.025 mm<br />
26 Max. axiale Belastung (dynamisch) 5.6 N<br />
27 Max. axiale Aufpresskraft (statisch) 110 N<br />
(statisch, Welle abgestützt) 1200 N<br />
28 Max. radiale Belastung, 5 mm ab Flansch 28 N<br />
Weitere Spezifikationen<br />
29 Polpaarzahl 1<br />
30 Anzahl Kollektorsegmente 13<br />
31 Motorgewicht 340 g<br />
Motordaten gemäss Tabelle sind Nenndaten.<br />
Erläuterungen zu den Ziffern Seite 47.<br />
Option<br />
Hohlwelle als Spezialausführung<br />
Vorgespannte Kugellager<br />
Betriebsbereiche <strong>Leg</strong>ende<br />
n [min -1<br />
]<br />
Dauerbetriebsbereich<br />
Unter Berücksichtigung der angegebenen thermischen<br />
Widerstände (Ziffer 17 und 18) und einer Umgebungstemperatur<br />
von 25°C wird bei dauernder<br />
Belastung die maximal zulässige Rotortemperatur<br />
erreicht = thermische Grenze.<br />
Kurzzeitbetrieb<br />
Der Motor darf kurzzeitig und wiederkehrend überlastet<br />
werden.<br />
Typenleistung<br />
maxon-Baukastensystem Übersicht Seite 16 - 21<br />
Planetengetriebe<br />
�32 mm<br />
0.75 - 4.5 Nm<br />
Seite 230<br />
Planetengetriebe<br />
�32 mm<br />
1.0 - 6.0 Nm<br />
Seite 231<br />
Planetengetriebe<br />
�32 mm<br />
8 Nm<br />
Seite 233<br />
Planetengetriebe<br />
�42 mm<br />
3 - 15 Nm<br />
Seite 235<br />
Empfohlene Elektronik:<br />
ADS 50/5 Seite 268<br />
ADS 50/10 269<br />
ADS_E 50/5 269<br />
ADS_E 50/10 269<br />
EPOS 24/5 286<br />
EPOS P 24/5 287<br />
MIP 50 289<br />
Hinweise 17<br />
Encoder MR<br />
256 - 1024 Imp.,<br />
3 Kanal<br />
Seite 251<br />
Encoder HED_ 5540<br />
500 Imp.,<br />
3 Kanal<br />
Seite 254 / 256<br />
DC-Tacho DCT<br />
�22 mm<br />
0.52 V<br />
Seite 263<br />
Bremse AB 28<br />
�28 mm<br />
24 VDC, 0.4 Nm<br />
Seite 300<br />
Ausgabe April 2007 / Änderungen vorbehalten maxon DC motor 81<br />
maxon DC motor
A.3 Data sheets 55<br />
Planetengetriebe GP 42 C �42 mm, 3 - 15 Nm<br />
Keramikversion<br />
Lagerprogramm<br />
Standardprogramm<br />
Bestellnummern<br />
Sonderprogramm (auf Anfrage)<br />
Getriebedaten<br />
203113 203115 203119 203120 203124 203129 203128 203133 203137 203141<br />
1 Untersetzung 3.5 : 1 12 : 1 26 : 1 43 : 1 81 : 1 156 : 1 150 : 1 285 : 1 441 : 1 756 : 1<br />
2 Untersetzung absolut 7<br />
/2<br />
49<br />
/4 26 343<br />
/8<br />
2197<br />
/27 156 2401<br />
/16<br />
15379<br />
/54 441 756<br />
3 Massenträgheitsmoment gcm2 14 15 9.1 15 9.4 9.1 15 15 14 14<br />
4 Max. Motorwellendurchmesser mm 10 10 8 10 8 8 10 10 10 10<br />
Bestellnummern 203114 203116 203121 203125 203130 203134 203138 203142<br />
1 Untersetzung 4.3 : 1 15 : 1 53 : 1 91 : 1 186 : 1 319 : 1 488 : 1 936 : 1<br />
2 Untersetzung absolut 13<br />
/3<br />
91<br />
/6<br />
637<br />
/12 91 4459<br />
/24<br />
637<br />
/2<br />
4394<br />
/9 936<br />
3 Massenträgheitsmoment gcm2 9.1 15 15 15 15 15 9.4 9.1<br />
4 Max. Motorwellendurchmesser mm 8 10 10 10 10 10 8 8<br />
Bestellnummern 203117 203122 203126 203131 203135 203139<br />
1 Untersetzung 19 : 1 66 : 1 113 : 1 230 : 1 353 :1 546 : 1<br />
2 Untersetzung absolut 169<br />
/9<br />
1183<br />
/18<br />
338<br />
/3<br />
8281<br />
/36<br />
28561<br />
/81 546<br />
3 Massenträgheitsmoment gcm2 9.4 15 9.4 15 9.4 14<br />
4 Max. Motorwellendurchmesser mm 8 10 8 10 8 10<br />
Bestellnummern 203118 203123 203127 203132 203136 203140<br />
1 Untersetzung 21 : 1 74 : 1 126 : 1 257 : 1 394 : 1 676 : 1<br />
2 Untersetzung absolut 21 147<br />
/2 126 1029<br />
/4 1183 /3 676<br />
3 Massenträgheitsmoment gcm2 14 15 14 15 15 9.1<br />
4 Max. Motorwellendurchmesser mm 10 10 10 10 10 8<br />
5 Stufenzahl 1 2 2 3 3 3 4 4 4 4<br />
6 Max. Dauerdrehmoment Nm 3.0 7.5 7.5 15.0 15.0 15.0 15.0 15.0 15.0 15.0<br />
7 kurzzeitig zulässiges Drehmoment Nm 4.5 11.3 11.3 22.5 22.5 22.5 22.5 22.5 22.5 22.5<br />
8 Max. Wirkungsgrad % 90 81 81 72 72 72 64 64 64 64<br />
9 Gewicht g 260 360 360 460 460 460 560 560 560 560<br />
10 Mittleres Getriebespiel unbelastet ° 0.3 0.4 0.4 0.5 0.5 0.5 0.5 0.5 0.5 0.5<br />
11 Getriebelänge L1 mm 40.9 55.4 55.4 69.9 69.9 69.9 84.4 84.4 84.4 84.4<br />
Gesamtlänge<br />
Gesamtlänge<br />
M 1:2<br />
Technische Daten<br />
Planetengetriebe geradeverzahnt<br />
Abtriebswelle rostfreier Stahl<br />
Abtriebswellenlagerung vorgespannte Kugellager<br />
Radialspiel, 12 mm ab Flansch max. 0.06 mm<br />
Axialspiel bei Axiallast < 5 N 0 mm<br />
>5N max.0.3mm<br />
Max. zulässige Axiallast 150 N<br />
Max. zulässige Aufpresskraft 300 N<br />
Drehsinn, Antrieb zu Abtrieb =<br />
Empfohlene Motordrehzahl < 8000 min -1<br />
Empfohlener Temperaturbereich -20 ... +100°C<br />
erweiterter Bereich als Option -35 ... +100°C<br />
Stufenzahl 1 2 3 4<br />
Max. zul. Radiallast,<br />
12 mm ab Flansch 120 N 150 N 150 N 150 N<br />
Kombination<br />
+ Motor Seite +Tacho Seite +BremseSeite = Motorlänge + Getriebelänge + (Tacho / Bremse) + Montageteile<br />
RE 35, 90 W 81 111.9 126.4 126.4 140.9 140.9 140.9 155.4 155.4 155.4 155.4<br />
RE 35, 90 W 81 MR 251 123.3 137.8 137.8 152.3 152.3 152.3 166.8 166.8 166.8 166.8<br />
RE 35, 90 W 81 HED_ 5540 254/256 132.9 147.4 147.4 161.9 161.9 161.9 176.4 176.4 176.4 176.4<br />
RE 35, 90 W 81 DCT 22 263 130.0 144.5 144.5 159.0 159.0 159.0 173.5 173.5 173.5 173.5<br />
RE 35, 90 W 81 AB 28 300 148.0 162.5 162.5 177.0 177.0 177.0 191.5 191.5 191.5 191.5<br />
RE 36, 70 W 82 112.2 126.7 126.7 141.2 141.2 141.2 155.7 155.7 155.7 155.7<br />
RE 36, 70 W 82 MR 251 123.6 138.1 138.1 152.6 152.6 152.6 167.1 167.1 167.1 167.1<br />
RE 36, 70 W 82 HED_ 5540 254/256 133.2 147.7 147.7 162.2 162.2 162.2 176.7 176.7 176.7 176.7<br />
RE 36, 70 W 82 DCT 22 263 130.3 144.8 144.8 159.3 159.3 159.3 173.8 173.8 173.8 173.8<br />
RE 40, 150 W 83 112.0 126.5 126.5 141.0 141.0 141.0 155.5 155.5 155.5 155.5<br />
RE 40, 150 W 83 MR 251 123.4 137.9 137.9 152.4 152.4 152.4 166.9 166.9 166.9 166.9<br />
RE 40, 150 W 83 HED_ 5540 254/256 132.7 147.2 147.2 161.7 161.7 161.7 176.2 176.2 176.2 176.2<br />
RE 40, 150 W 83 HEDL 9140 259 166.1 180.6 180.6 195.1 195.1 195.1 209.6 209.6 209.6 209.6<br />
RE 40, 150 W 83 AB 28 300 148.1 162.6 162.6 177.1 177.1 177.1 191.6 191.6 191.6 191.6<br />
RE 40, 150 W 83 AB 28 301 156.1 170.6 170.6 185.1 185.1 185.1 199.6 199.6 199.6 199.6<br />
RE 40, 150 W 83 HED_ 5540 254/256 AB 28 300 165.2 179.7 179.7 194.2 194.2 194.2 208.7 208.7 208.7 208.7<br />
RE 40, 150 W 83 HEDL 9140 259 AB 28 301 176.6 191.1 191.1 205.6 205.6 205.6 220.1 220.1 220.1 220.1<br />
EC 40, 120 W 163 111.0 125.5 125.5 140.0 140.0 140.0 154.5 154.5 154.5 154.5<br />
EC 40, 120 W 163 HED_ 5540 255/257 129.4 143.9 143.9 158.4 158.4 158.4 172.9 172.9 172.9 172.9<br />
EC 40, 120 W 163 Res 26 264 137.6 152.1 152.1 166.6 166.6 166.6 181.1 181.1 181.1 181.1<br />
EC 40, 120 W 163 AB 28 300 141.8 156.3 156.3 170.8 170.8 170.8 185.3 185.3 185.3 185.3<br />
EC 45, 150 W 164 152.2 166.7 166.7 181.2 181.2 181.2 195.7 195.7 195.7 195.7<br />
EC 45, 150 W 164 HEDL 9140 259 167.8 182.3 182.3 196.8 196.8 196.8 211.3 211.3 211.3 211.3<br />
EC 45, 150 W 164 Res 26 264 152.2 166.7 166.7 181.2 181.2 181.2 195.7 195.7 195.7 195.7<br />
EC 45, 150 W 164 AB 28 301 159.6 174.1 174.1 188.6 188.6 188.6 203.1 203.1 203.1 203.1<br />
Ausgabe April 2007 / Änderungen vorbehalten maxon gear 235<br />
maxon gear
A.3 Data sheets 56<br />
maxon DC motor<br />
RE 25 �25 mm, Graphitbürsten, 20 Watt<br />
Lagerprogramm<br />
Standardprogramm<br />
Sonderprogramm (auf Anfrage)<br />
Bestellnummern<br />
gemäss Massbild 118749 118750 118751 118752 118753 118754 118755 118756 118757<br />
Wellenlänge 15.7 gekürzt auf 4 mm 302002 302003 302004 302005 302006 302007 302001 302008 302009<br />
Motordaten<br />
Werte bei Nennspannung<br />
1 Nennspannung V 9.0 15.0 18.0 24.0 30.0 42.0 48.0 48.0 48.0<br />
2 Leerlaufdrehzahl min-1 10000 9650 10200 9550 9860 11100 10300 8230 5050<br />
3 Leerlaufstrom mA 110 60.7 53.9 36.9 30.5 25.2 20.1 15.2 8.51<br />
4 Nenndrehzahl min-1 8980 8470 8890 8360 8680 9950 9190 7070 3870<br />
5 Nennmoment (max. Dauerdrehmoment) mNm 11.1 20.6 23.1 26.7 27.2 27.6 28.4 29.4 30.8<br />
6 Nennstrom (max. Dauerbelastungsstrom) A 1.50 1.50 1.47 1.17 0.983 0.799 0.667 0.548 0.352<br />
7 Anhaltemoment mNm 244 237 233 257 263 299 280 222 136<br />
8 Anlaufstrom A 30.7 16.6 14.3 11.0 9.21 8.39 6.38 4.03 1.52<br />
9 Max. Wirkungsgrad<br />
Kenndaten<br />
% 77 83 84 86 86 88 88 87 85<br />
10 Anschlusswiderstand � 0.293 0.902 1.26 2.19 3.26 5.00 7.53 11.9 31.6<br />
11 Anschlussinduktivität mH 0.0275 0.0882 0.115 0.238 0.353 0.551 0.832 1.31 3.48<br />
12 Drehmomentkonstante mNm A-1 7.97 14.3 16.3 23.4 28.5 35.7 43.8 55.0 89.7<br />
13 Drehzahlkonstante min-1 V-1 1200 669 585 407 335 268 218 173 106<br />
14 Kennliniensteigung min-1 mNm-1 44.1 42.3 45.3 38.1 38.2 37.5 37.4 37.6 37.5<br />
15 Mechanische Anlaufzeitkonstante ms 5.36 4.58 4.49 4.28 4.20 4.13 4.11 4.10 4.09<br />
16 Rotorträgheitsmoment gcm2 11.6 10.3 9.45 10.7 10.5 10.5 10.5 10.4 10.4<br />
Spezifikationen<br />
Thermische Daten<br />
17 Therm. Widerstand Gehäuse-Luft 14 KW -1<br />
18 Therm. Widerstand Wicklung-Gehäuse 3.1 KW -1<br />
19 Therm. Zeitkonstante der Wicklung 12.4 s<br />
20 Therm. Zeitkonstante des Motors 910 s<br />
21 Umgebungstemperatur -20 ... +100°C<br />
22 Max. Wicklungstemperatur +125°C<br />
Mechanische Daten (Kugellager)<br />
23 Grenzdrehzahl 14000 min -1<br />
24 Axialspiel 0.05 - 0.15 mm<br />
25 Radialspiel 0.025 mm<br />
26 Max. axiale Belastung (dynamisch) 3.2 N<br />
27 Max. axiale Aufpresskraft (statisch) 64 N<br />
(statisch, Welle abgestützt) 270 N<br />
28 Max. radiale Belastung, 5 mm ab Flansch 16 N<br />
Weitere Spezifikationen<br />
29 Polpaarzahl 1<br />
30 Anzahl Kollektorsegmente 11<br />
31 Motorgewicht 130 g<br />
Motordaten gemäss Tabelle sind Nenndaten.<br />
Erläuterungen zu den Ziffern Seite 47.<br />
Option<br />
Vorgespannte Kugellager<br />
Betriebsbereiche <strong>Leg</strong>ende<br />
n [min -1<br />
]<br />
Dauerbetriebsbereich<br />
Unter Berücksichtigung der angegebenen thermischen<br />
Widerstände (Ziffer 17 und 18) und einer Umgebungstemperatur<br />
von 25°C wird bei dauernder<br />
Belastung die maximal zulässige Rotortemperatur<br />
erreicht = thermische Grenze.<br />
Kurzzeitbetrieb<br />
Der Motor darf kurzzeitig und wiederkehrend überlastet<br />
werden.<br />
Typenleistung<br />
maxon-Baukastensystem Übersicht Seite 16 - 21<br />
Planetengetriebe<br />
�26 mm<br />
0.5 - 2.0 Nm<br />
Seite 226<br />
Planetengetriebe<br />
�32 mm<br />
0.4 - 2.0 Nm<br />
Seite 228<br />
Planetengetriebe<br />
�32 mm<br />
0.75 - 6.0 Nm<br />
Seite 229 / 231<br />
Empfohlene Elektronik:<br />
LSC 30/2 Seite 268<br />
ADS 50/5 268<br />
ADS_E 50/5 269<br />
EPOS 24/5 286<br />
EPOS P 24/5 287<br />
MIP 10 289<br />
Hinweise 17<br />
M 1:2<br />
Encoder MR<br />
128 - 1000 Imp.,<br />
3 Kanal<br />
Seite 250<br />
Encoder Enc<br />
22 mm<br />
100 Imp., 2 Kanal<br />
Seite 252<br />
Encoder HED_ 5540<br />
500 Imp.,<br />
3 Kanal<br />
Seite 254 / 256<br />
DC-Tacho DCT<br />
�22 mm,<br />
0.52 V<br />
Seite 263<br />
Bremse AB 28<br />
�28 mm<br />
24 VDC, 0.4 Nm<br />
Seite 300<br />
78 maxon DC motor Ausgabe April 2007 / Änderungen vorbehalten
A.3 Data sheets 57<br />
Planetengetriebe GP 32 C �32 mm, 1.0 - 6.0 Nm<br />
Keramikversion<br />
Lagerprogramm<br />
Standardprogramm<br />
Bestellnummern<br />
Sonderprogramm (auf Anfrage)<br />
Getriebedaten<br />
166930 166933 166938 166939 166944 166949 166954 166959 166962 166967 166972 166977<br />
1 Untersetzung 3.7 : 1 14 : 1 33 : 1 51 : 1 111 : 1 246 : 1 492 : 1 762 : 1 1181 : 1 1972 : 1 2829 : 1 4380 : 1<br />
2 Untersetzung absolut 26<br />
/7<br />
676<br />
/49<br />
529<br />
/16<br />
17576<br />
/343<br />
13824 421824<br />
/125 /1715 86112<br />
/175<br />
19044 10123776<br />
/25 /8575 8626176 /4375 495144<br />
/175<br />
109503<br />
/25<br />
3 Max. Motorwellendurchmesser mm 6 6 3 6 4 4 3 3 4 4 3 3<br />
Bestellnummern 166931 166934 166940 166945 166950 166955 166960 166963 166968 166973 166978<br />
1 Untersetzung 4.8 : 1 18 : 1 66 : 1 123 : 1 295 : 1 531 : 1 913 : 1 1414 : 1 2189 : 1 3052 : 1 5247 : 1<br />
2 Untersetzung absolut 24<br />
/5<br />
624<br />
/35<br />
16224<br />
/245<br />
6877<br />
/56<br />
101062<br />
/343<br />
331776<br />
/625<br />
36501 2425488<br />
/40 /1715 536406 1907712<br />
/245 /625 839523<br />
/160<br />
3 Max. Motorwellendurchmesser mm 4 4 4 3 3 4 3 3 3 3 3<br />
Bestellnummern 166932 166935 166941 166946 166951 166956 166961 166964 166969 166974 166979<br />
1 Untersetzung 5.8 : 1 21 : 1 79 : 1 132 : 1 318 : 1 589 : 1 1093 : 1 1526 : 1 2362 : 1 3389 : 1 6285 : 1<br />
2 Untersetzung absolut 23<br />
/4 299 /14<br />
3887 /49 3312 /25<br />
389376 /1225 20631<br />
/35 279841 9345024<br />
/256 /6125 2066688 /875 474513 6436343<br />
/140 /1024<br />
3 Max. Motorwellendurchmesser mm 3 3 3 3 4 3 3 4 3 3 3<br />
Bestellnummern 166936 166942 166947 166952 166957 166965 166970 166975<br />
1 Untersetzung 23 : 1 86 : 1 159 : 1 411 : 1 636 : 1 1694 : 1 2548 : 1 3656 : 1<br />
2 Untersetzung absolut 576<br />
/25<br />
14976<br />
/175<br />
1587<br />
/10<br />
359424<br />
/875<br />
79488<br />
/125<br />
1162213 /686 7962624 /3125 457056<br />
/125<br />
3 Max. Motorwellendurchmesser mm 4 4 3 4 3 3 4 3<br />
Bestellnummern 166937 166943 166948 166953 166958 166966 166971 166976<br />
1 Untersetzung 28 : 1 103 : 1 190 : 1 456 : 1 706 : 1 1828 : 1 2623 : 1 4060 : 1<br />
2 Untersetzung absolut 138<br />
/5<br />
3588<br />
/35<br />
12167<br />
/64<br />
89401<br />
/196<br />
158171<br />
/224<br />
2238912 /1225 2056223 /784 3637933 /896<br />
3 Max. Motorwellendurchmesser mm 3 3 3 3 3 3 3 3<br />
4 Stufenzahl 1 2 2 3 3 4 4 4 5 5 5 5<br />
5 Max. Dauerdrehmoment Nm 1 3 3 6 6 6 6 6 6 6 6 6<br />
6 kurzzeitig zulässiges Drehmoment Nm 1.25 3.75 3.75 7.5 7.5 7.5 7.5 7.5 7.5 7.5 7.5 7.5<br />
7 Max. Wirkungsgrad % 80 75 75 70 70 60 60 60 50 50 50 50<br />
8 Gewicht g 118 162 162 194 194 226 226 226 258 258 258 258<br />
9 Mittleres Getriebespiel unbelastet ° 0.7 0.8 0.8 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0<br />
10 Massenträgheitsmoment gcm2 1.5 0.8 0.8 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7<br />
11 Getriebelänge L1 mm 26.4 36.3 36.3 43.0 43.0 49.7 49.7 49.7 56.4 56.4 56.4 56.4<br />
Gesamtlänge<br />
Gesamtlänge<br />
M 1:2<br />
Technische Daten<br />
Planetengetriebe geradeverzahnt<br />
Abtriebswelle rostfreier Stahl<br />
Wellendurchmesser als Option 8 mm<br />
Abtriebswellenlagerung Kugellager<br />
Radialspiel, 5 mm ab Flansch max. 0.14 mm<br />
Axialspiel max. 0.4 mm<br />
Max. zul. Radiallast, 10 mm ab Flansch 140 N<br />
Max. zulässige Axiallast 120 N<br />
Max. zulässige Aufpresskraft 120 N<br />
Drehsinn, Antrieb zu Abtrieb =<br />
Empfohlene Motordrehzahl < 8000 min -1<br />
Empfohlener Temperaturbereich -20 ... +100°C<br />
erweiterter Bereich als Option -35 ... +100°C<br />
Option: Geräuschreduzierte Ausführung<br />
Kombination<br />
+ Motor Seite +Tacho/BremseSeite Gesamtlänge [mm] = Motorlänge + Getriebelänge + (Tacho / Bremse) + Montageteile<br />
RE 25, 10 W 76 81.0 90.9 90.9 97.6 97.6 104.3 104.3 104.3 111.0 111.0 111.0 111.0<br />
RE 25, 10 W 76 MR 250 92.0 101.9 101.9 108.6 108.6 115.3 115.3 115.3 122.0 122.0 122.0 122.0<br />
RE 25, 10 W 76 Enc 22 252 95.1 105.0 105.0 111.7 111.7 118.4 118.4 118.4 125.1 125.1 125.1 125.1<br />
RE 25, 10 W 76 HED_ 5540 254/256 101.8 111.7 111.7 118.4 118.4 125.1 125.1 125.1 131.8 131.8 131.8 131.8<br />
RE 25, 10 W 76 DCT 22 263 103.3 113.2 113.2 119.9 119.9 126.6 126.6 126.6 133.3 133.3 133.3 133.3<br />
RE 25, 20 W 77 69.5 79.4 79.4 86.1 86.1 92.8 92.8 92.8 99.5 99.5 99.5 99.5<br />
RE 25, 20 W 78 81.0 90.9 90.9 97.6 97.6 104.3 104.3 104.3 111.0 111.0 111.0 111.0<br />
RE 25, 20 W 78 MR 250 92.0 101.9 101.9 108.6 108.6 115.3 115.3 115.3 122.0 122.0 122.0 122.0<br />
RE 25, 20 W 78 Enc 22 252 95.1 105.0 105.0 111.7 111.7 118.4 118.4 118.4 125.1 125.1 125.1 125.1<br />
RE 25, 20 W 78 HED_ 5540 254/256 101.8 111.7 111.7 118.4 118.4 125.1 125.1 125.1 131.8 131.8 131.8 131.8<br />
RE 25, 20 W 78 DCT 22 263 103.3 113.2 113.2 119.9 119.9 126.6 126.6 126.6 133.3 133.3 133.3 133.3<br />
RE 25, 20 W 78 HED_5540 / AB 28 300 132.2 142.1 142.1 148.8 148.8 155.5 155.5 155.5 162.2 162.2 162.2 162.2<br />
RE 26, 18 W 79 85.3 95.2 95.2 101.9 101.9 108.6 108.6 108.6 115.3 115.3 115.3 115.3<br />
RE 26, 18 W 79 MR 250 96.3 106.2 106.2 112.9 112.9 119.6 119.6 119.6 126.3 126.3 126.3 126.3<br />
RE 26, 18 W 79 Enc 22 252 102.7 112.6 112.6 119.3 119.3 126.0 126.0 126.0 132.7 132.7 132.7 132.7<br />
RE 26, 18 W 79 HED_ 5540 254/256 103.7 113.6 113.6 120.3 120.3 127.0 127.0 127.0 133.7 133.7 133.7 133.7<br />
RE 26, 18 W 79 DCT 22 263 106.3 116.2 116.2 122.9 122.9 129.6 129.6 129.6 136.3 136.3 136.3 136.3<br />
RE 30, 60 W 80 94.5 104.4 104.4 111.1 111.1 117.8 117.8 117.8 124.5 124.5 124.5 124.5<br />
RE 30, 60 W 80 MR 251 105.9 115.8 115.8 122.5 122.5 129.2 129.2 129.2 135.9 135.9 135.9 135.9<br />
RE 35, 90 W 81 97.4 107.3 107.3 114.0 114.0 120.7 120.7 120.7 127.4 127.4 127.4 127.4<br />
RE 35, 90 W 81 MR 251 108.8 118.7 118.7 125.4 125.4 132.1 132.1 132.1 138.8 138.8 138.8 138.8<br />
RE 35, 90 W 81 HED_ 5540 254/256 118.4 128.3 128.3 135.0 135.0 141.7 141.7 141.7 148.4 148.4 148.4 148.4<br />
RE 35, 90 W 81 DCT 22 263 115.5 125.4 125.4 132.1 132.1 138.8 138.8 138.8 145.5 145.5 145.5 145.5<br />
RE 35, 90 W 81 AB 28 300 133.5 143.4 143.4 150.1 150.1 156.8 156.8 156.8 163.5 163.5 163.5 163.5<br />
RE 35, 90 W 81 HEDS 5540 / AB 28 254/300 150.6 160.5 160.5 167.2 167.2 173.9 173.9 173.9 180.6 180.6 180.6 180.6<br />
RE 36, 70 W 82 97.7 107.6 107.6 114.3 114.3 121.0 121.0 121.0 127.7 127.7 127.7 127.7<br />
RE 36, 70 W 82 MR 251 109.1 119.0 119.0 125.7 125.7 132.4 132.4 132.4 139.1 139.1 139.1 139.1<br />
RE 36, 70 W 82 HED_ 5540 254/256 118.7 128.6 128.6 135.3 135.3 142.0 142.0 142.0 148.7 148.7 148.7 148.7<br />
RE 36, 70 W 82 DCT 22 263 115.8 125.7 125.7 132.4 132.4 139.1 139.1 139.1 145.8 145.8 145.8 145.8<br />
Ausgabe April 2007 / Änderungen vorbehalten 231<br />
maxon gear
A.3 Data sheets 58<br />
Druckvorschau<br />
http://www.federnshop.com/Datenblaetter/DruckVorschau.aspx?ArtNr=D-357<br />
Page 1 of 1<br />
17.12.2007
A.3 Data sheets 59<br />
Willkommen in unserem Federauswahlprogramm und Online Shop - Federntechnik Knör...<br />
Federtyp / Werkstoff: Kriterien: von: bis:<br />
Schenkelfeder Mt - max. Drehmoment (Nmm) bei α 10000<br />
suchen<br />
<br />
Direktsuche nach Artnr.:<br />
suchen<br />
(% = Platzhalter/Wildcard)<br />
<br />
<br />
<br />
<br />
Auswahl - Ergebnisliste: Anzahl gefundener Artikel: 16<br />
Artnr. Werkst. Menge Warenkorb Details CAD-3D d Di De n Lk Grad α Mt Rmr Dd Dh Ls Bild<br />
M45L20 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.00 29.41 75.45 10019.72 132.79 25.50 39.00 120.00 A<br />
M45R20 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.00 29.41 75.45 10019.72 132.79 25.50 39.00 120.00 A<br />
M45L21 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.25 30.54 79.23 10019.72 126.47 25.50 39.00 120.00 B<br />
M45R21 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.25 30.54 79.23 10019.72 126.47 25.50 39.00 120.00 B<br />
M45L22 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.50 31.67 83.00 10019.72 120.72 25.50 39.00 120.00 C<br />
M45R22 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.50 31.67 83.00 10019.72 120.72 25.50 39.00 120.00 C<br />
M45L23 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.75 32.81 86.77 10019.72 115.47 25.50 39.00 120.00 D<br />
M45R23 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 5.75 32.81 86.77 10019.72 115.47 25.50 39.00 120.00 D<br />
M45L32 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 7.50 40.72 113.18 10019.72 88.53 25.50 39.00 120.00 C<br />
M45R32 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 7.50 40.72 113.18 10019.72 88.53 25.50 39.00 120.00 C<br />
M45L40 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 8.00 42.99 120.73 10019.72 83.00 25.50 39.00 120.00 A<br />
M45R40 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 8.00 42.99 120.73 10019.72 83.00 25.50 39.00 120.00 A<br />
M45L50 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 11.00 56.56 166.00 10019.72 60.36 25.50 39.00 120.00 A<br />
M45R50 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 11.00 56.56 166.00 10019.72 60.36 25.50 39.00 120.00 A<br />
M45L60 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 15.00 74.66 226.36 10019.72 44.26 25.50 39.00 120.00 A<br />
M45R60 1.4310 0 in Korb Daten CAD-3D 4.50 27.50 36.50 15.00 74.66 226.36 10019.72 44.26 25.50 39.00 120.00 A<br />
Home | Shop | AGB | Impressum | Lieferbedingungen | Anmelden | Login | Warenkorb (0) © 2007 by FK-Knörzer GmbH; erstellt von VKNet<br />
http://shop.knoerzer.eu/shop.php<br />
C/CE<br />
W.Korb<br />
Login<br />
Page 1 of 1<br />
27.12.2007
A.3 Data sheets 60<br />
���<br />
��� �������� ������ � ���������������<br />
������������������� �� ��������������<br />
������ ����������� �������� ��������<br />
������������ ���<br />
����������������� ���� ���<br />
������������ ���<br />
����������������� ���� ���<br />
������������ ���<br />
����������������� ���� ���<br />
������������ ���<br />
����������������� ���� ���<br />
���������� ����� �������� ��������������<br />
����� �������� ����� �������� ������� �����������<br />
�� ����� �������� ������ � �������<br />
�������������� ������ �� ������� � ����<br />
��� ���� ������ �������������<br />
���� ������������ ����<br />
����� ��������� ����������<br />
��������� ����� � ��������<br />
������� ��� ������<br />
���� ��� � ����������� � �� � � � � � � ����<br />
�� ����� ��� �� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ����� ��� �� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ����� ��� �� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ����� ����� �� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ����� ��� �� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ���� ���� �� ��� �� ���� �� ��� ��<br />
�� ����� ����� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ����� ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ��� �<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ����� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ����� ��<br />
�� ������ ����� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ����� ����� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ����� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ����� �� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
�� ������ ��� �� ���� ���� �� ��� �� ���� �� ���� �<br />
�� ���� ���� �� ��� �� ���� �� ���� ��<br />
� �� ����� ����������� � ����<br />
����������� ��� � ���<br />
��������� ��� ���������<br />
������������ ��� � ���<br />
������� ���������� ���<br />
�����<br />
���
A.3 Data sheets 61<br />
Katalog<br />
Antriebselemente und Normteile<br />
Allgemeine Normteile<br />
Auszugschienen<br />
Buchsen<br />
Federnde Druckstücke<br />
Gewindestangen, Muttern, Schrauben<br />
Gabelgelenke,Gabelköpfe und ES-Bo<br />
Gelenkköpfe DIN 648 und Gelenklage<br />
Griffe, Knöpfe und Bedienteile<br />
Höhenverstellbare Elemente<br />
Hub- und Verstellsysteme<br />
Innenzahnkränze<br />
Kantenschutzprofile<br />
Kegelräder<br />
Keilriemen und Keilriemenscheiben<br />
Keilwellen und Keilnaben<br />
Kettenräder<br />
Kleinstoßdämpfer und Zubehör<br />
Kugelrollen<br />
Kupplungen<br />
Ausgleichskupplungen KA<br />
Ausgleichskupplungen LA<br />
Zahn-Kupplungen BOS<br />
Zahn-Kupplungen BOZ<br />
Zahn-Kupplungen BW<br />
Drehstarre Kupplungen HB<br />
Drehstarre Kupplungen HF<br />
Drehstarre Kupplungen HFD mit D<br />
Drehstarre Kupplungen HU<br />
Drehstarre Kupplungen HZ<br />
Drehstarre Kupplungen HZD mit D<br />
Elastische Kupplungen DX<br />
Elastische Kupplungen ME<br />
Elastische Kupplungen MU<br />
Elastische Kupplungen RN<br />
Elastische Kupplungen RN Edels<br />
http://www.maedler.ch/katalog_ch/katalog/index.htm<br />
Ausgleichskupplungen LA<br />
Artikelnummer: 60293600<br />
Staffelpreise in CHF pro Stück, bei Abnahme von:<br />
1 5 10 25 50 100<br />
28.97 27.34 25.76 24.13 22.53 21.94<br />
Artikeldaten<br />
Werkstoff 2)<br />
Max. Betriebsdrehmomente Nm 10,00<br />
B mm 8<br />
L1 mm 38,10<br />
L2 mm 11,00<br />
D1 mm 25,40<br />
Winkelversatz Grad 7<br />
Parallelversatz mm 0,38<br />
Gewicht g 44<br />
Zurück zur Artikelansicht<br />
Page 1 of 1<br />
04.04.2008
A.4 Parts list 62<br />
A.4 Parts list
References 63<br />
References<br />
[1] A.A. Frank. Automatic control systems for legged locomotion. Technical<br />
Report USCEE Report No. 273, University of Southern California, Los<br />
Angeles, 1966.<br />
[2] R.B. McGhee. Finite state control of quadruped locomotion. In Proceedings<br />
of the Second International Symposium on External Control of Human<br />
Extremities, Dubrovnik, Yugoslavia, 1966.<br />
[3] K. Matsuoka. A model of repetitive hopping movements in man. Proceed-<br />
ings of the Fifth World Congress on Theory of Machines and Mechanisms,<br />
International Federation for Information Processing, 1979.<br />
[4] M.H. Raibert. Trotting, pacing and bounding by a quadruped robot. Jour-<br />
nal of Biomechanics, 23(Supp. 1):79–98, 1990.<br />
[5] B.W. Lilly. Design and analysis of a mechanically coordinated, dynami-<br />
cally stable, quadruped trotting machine. Master thesis, The Ohio State<br />
University, 1986.<br />
[6] J. Furusho, S. Akihito, S. Masamichi, and K. Eichi. Realization of bounce<br />
gait in a quadruped robot <strong>with</strong> articular-joint-type- legs. Proceedings of<br />
the IEEE International Conference on <strong>Robotic</strong>s and Automation, pages<br />
697–702, 1995.<br />
[7] H. Kimura, S. Akyama, and K. Sakurama. Realization of dynamic walking<br />
and running of the quadruped using neural oscillator. Autonomous Robots,<br />
7:247–258, 1999.<br />
[8] H. Kimura, Y. Fukuoka, Y. Hada, and K. Takase. Three-dimensional adap-<br />
tive dynamic walking of a quadruped - rolling motion feedback to cpg’s con-<br />
trolling pitching motion. Proceedings of IEEE International Conference on<br />
<strong>Robotic</strong>s and Automation, pages 2228–2233, 2002.<br />
[9] J.P. Schmiedeler. The Mechanics of and <strong>Robotic</strong> Design for Quadrupedal<br />
Galloping. PhD thesis, The Ohio State University.<br />
[10] J.P. Schmiedeler, D.W. Marhefka, D.E. Orin, and K.J. Waldron. A study<br />
of quadruped gallops. NSF Design, Service and Manufacturing Grantees<br />
and Research Conference, 2001.<br />
[11] J.G. Nicol, L.R. Palmer, and K.J. Waldron. Design of a leg system for<br />
quadruped gallop. Proceedings of the 11th World Congress in Mechanism<br />
and Machine Science, 2003.
References 64<br />
[12] D. Papadopoulos and M. Buehler. Stable running in a quadruped robot<br />
<strong>with</strong> compliant legs. Proceedings of the IEEE International Conference on<br />
<strong>Robotic</strong>s and Automation, 2000.<br />
[13] R. Ringrose. Self-stabilizing running. 1996.<br />
[14] J.A. Smith and I. Poulakakis. Rotary gallop in the untethered quadrupedal<br />
robot scout ii. Proceedings of the International Conference on Intelligent<br />
Robots and Systems (IROS), 2004.<br />
[15] D.W. Robinson, J.E. Pratt, D.J. Paluska, and G.A. Pratt. Series elastic<br />
actuator development for a biomimetic walking robot. AIM, 1:561–568,<br />
1999.<br />
[16] D. Paluska and H. Herr. The effect of series elasticity on actuator power<br />
and work output: Implications for robotic and prosthetic joint design.<br />
<strong>Robotic</strong>s and Autonomous Systems, 54:667–673, 2006.<br />
[17] G. Mennitto. Compliant articulated robot leg <strong>with</strong> antagonistic ladd ac-<br />
tuation. Master’s thesis, McGill University, Montreal, Canada, February<br />
1995.<br />
[18] J.M. Remic. Prototype leg design for a quadruped robot application. Mas-<br />
ter’s thesis, The Ohio State University, USA, 2005.<br />
[19] B.T. Knox. Evaluation of a prototype series-complient hopping leg for<br />
biped robot applications. Bachelor’s thesis, The Ohio State University,<br />
USA, 2007.<br />
[20] S. Curran. Analysis and optimization of a jump for a prototype leg wieth<br />
series-elastic actuation. Master’s thesis, The Ohio State University, USA,<br />
2007.<br />
[21] M. Toepfer. Mechanical design of a pantographic leg for dynamic walking<br />
<strong>with</strong> a quadruped robot. Bachelor’s thesis, ETH Zurich, Switzerland, July<br />
2007.<br />
[22] William F. Ganong. Review of Medical Physiology. McGraw-Hill Compa-<br />
nies, 21st edition edition, 2003.<br />
[23] www.federnshop.de. Gutekunst federn.<br />
[24] www.maedler.ch. Maedler.<br />
[25] www.kanya.de. Kanya ag.
References 65<br />
[26] http://kowalski.sf4.eu/addt/2kreise.html. Intersection of two circles. 2008.