Submitted version of the thesis - Airlab, the Artificial Intelligence ...

48 Chapter 4. Control
distance to the target position. Each distance is divided by a constant, then
multiplied with the multiplier defined as FAST, NORMAL and CLOSE.
Then another constant which is the minimum value that can initiate the
motors movement is added. Hence, the mapping that each value set by the
PWM signal generates a motion, is achieved. The multipliers are defined
according to the distance of the robot to the target object. For the distances
between 1200 mm and greater, the corresponding multiplier is classified as
FAST, between 1200 mm and 750 mm the multiplier is NORMAL, and for
the distances between 750 mm and 300 mm it is CLOSE. By defining such
control mechanism as FAST, NORMAL, CLOSE we are able to create three
different speed limits for the motors, which are all mapped from 60% to 85%
of the duty cycle where the motion is visible. As an example case, when
the robot detects a target at 3 meters, the motor speeds will be set with
multiplier FAST until the distance to the target reaches 1.2 meters. When
reached to 1.2 meters, the motor speed will be set with multiplier NORMAL
until the detected target distance is less than 700 millimeters. And finally,
the motors speeds will be set with multiplier CLOSE from 700 mm until the
target is reached. This mechanism allows to achieve more precision in the
motion, since the mapping guarantees that each motor speed set by PWM
will generate areasonablemotion. Moreover, insteadof tryingtomapall the
speeds uniquely, indeed we are trying to map only one third of the speeds,
from the predefined speed ranges to 60%-85% of the PWM value in the duty
cycle. Even though the robot goes to target with an unexpected path, the
path is corrected as the robot comes closer since the slow motor speeds are
more precise and easy to achieve.
Even though the solution is not working perfectly, the robot is able to
go to the target successfully almost every time. The travel time and number
of movements in order to reach the target are not optimal, but this result
should be regarded as normal since we are not using any encoder to detect
the movement in the motors.
The need for an optimization of the PWM signal is also related to the
result of the miscalculation of the required torque. We selected motors that
do not have enough torque power for our robot. This brought the problem
of the arrangement of the PWM signal in order to run the motors. Since
the motors have less torque than we expected, almost 50% of the PWM is
wasted. First trials to optimize this inefficiency have been made by changing
the frequency of the PWM to a higher or a lower frequency, but these
trials did not give good results. Later, we found the proper configuration,