06.02.2013 Views

Underwater Robots - Gianluca Antonelli.pdf

Underwater Robots - Gianluca Antonelli.pdf

Underwater Robots - Gianluca Antonelli.pdf

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

8.4 External Force Control 207<br />

be fulfilled along with the primary task as long as they do not conflict. For<br />

example, we can ask the system not to change the vehicle orientation, or not<br />

to use the vehicle at all as long asthe manipulator is working in adexterous<br />

posture.<br />

Let us define x p as the primary task vector and x s as the secondary task<br />

vector. The frame in which they are defined depends onthe variables weare<br />

interested in: if the primary task is the position of the end-effector it should<br />

be normally defined in the inertial frame.<br />

The task priority inverse kinematics algorithm is based onthe following<br />

update law [78]<br />

ζ d = J # p [ ˙x p,d + Λ p ( x p,d − x p )] +<br />

+(I − J # p J p ) J # s [ ˙x s,d + Λ s ( x s,d − x s )] , (8.7)<br />

where J p and J s are the configuration-dependent primary and secondary<br />

task Jacobians respectively, the symbol # denotes any kind of matrix inversion<br />

(e.g. Moore-Penrose), and Λ p and Λ s are positive definite matrices. It<br />

must be noted that x p , x s and the corresponding Jacobians J p , J s are functions<br />

of η d and q d .The vector ζ d is defined in the body-fixed frame and a<br />

suitable integration must be applied to obtain the desired positions: η d , q d<br />

(see (2.58)).<br />

Let us assume that the primary task is the end-effector position that<br />

implies that J p = J pos. Tointroduce aforce control action (8.7) is modified<br />

as follow: the vector ˙x c is given as areference value to the IK algorithm and<br />

the reference system velocities are computed as:<br />

where<br />

ζ d = J # p [ ˙x p,d + ˙x c + Λ p ( x p,d − x p )] +<br />

+(I − J # p J p ) J # s [ ˙x s,d + Λ s ( x s,d − x s )] , (8.8)<br />

˙x c = k f,p ˜ f e − k f,v ˙ f e + k f,i<br />

� t<br />

0<br />

˜f e ( σ ) dσ (8.9)<br />

being ˜ f e = f e,d − f e the force error. The direction in which force control is<br />

expected are included in the primary task vector, e.g., if the task requires to<br />

exert aforce along z ,the primary task includes the z component of x .<br />

8.4.2 Stability Analysis<br />

Pre-multiplying (8.8) by J p ∈ IR 3 × (6+n ) we obtain:<br />

J p ζ d = ˙x p,d + ˙x c + Λ p ( x p,d − x p ) . (8.10)<br />

Let us consider aregulation problem, i.e., the reference force f e,d and<br />

the desired primary task x p,d are constant. In addition, let us assume that<br />

f e,d ∈R( K ), where K is the stiffness matrix defined in (2.77), and that, as

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!