Mobile Robotics, 2nd Edition

3Model-free Control

When we implement a controller for a robot and perform the initial tests we rarely succeed on the first try, which leads us to the problem of debugging. It might be that the compass is subject to electromagnetic disturbances, that it is placed upside-down, that there is a unit conversion problem in the sensors, that the motors are saturated or that there is a sign problem in the equations of the controller. The problem of debugging is a complex one and it is wise to respect the continuity principle: each step in the construction of the robot must be of reasonable size and has to be validated before pursuing construction. Thus, for a robot, it is desirable to implement a simple intuitive controller that is easy to debug before setting up a more advanced one. This principle cannot always be applied. However, if we have a good a priori understanding of the control law to apply, then such a continuity principle can be followed. Among mobile robots for which a pragmatic controller can be imagined, we can distinguish at least two subclasses:

• – vehicle-robots: these are systems built by man to be controlled by man such as the bicycle, the sailboat, the car, etc. We will try to copy the control law used by humans and transform it into an algorithm;
• – biomimetic robots: these robots are inspired by the movement of human beings. We have been able to observe them for long periods of time and deduce the strategy developed by nature to design its control law. This is ...