ECAI 2004 Conference Paper

[full paper]

Learning Model Free Motor Control

Alejandro Agostini, Enric Celaya

Some robotic tasks, like the control of leg movements for a walking robot, require an accurate control to follow the desired trajectory in the presence of unforeseen external disturbances and variations in the dynamic parameters of the system. To solve this problem with classical control techniques such as PID, a precise tuning of control parameters is necessary, and they must be readjusted when the system working conditions vary. This burden can be avoided using a learning process that automatically learns the appropriate control law and adapts to ongoing system variations. A drawback of many learning systems is that they are not effective for non-toy problems, because of the large amount of experiences they require in the learning task. In this paper we present the results obtained with a categorization and learning algorithm able to perform efficient generalization of the observed situations, and learn accurate control policies in a short time without any previous knowledge of the plant and without the need of any kind of traditional control technique. Its performance is evaluated on the trajectory tracking control with simulated DC motors and compared with PID controls specifically tuned for the same problems.

Keywords: CATEGORIZATION AND LEARNING, REINFORCEMENT LEARNING, TRAJECTORY TRACKING CONTROL, ROBOT LOCOMOTION

Citation: Alejandro Agostini, Enric Celaya: Learning Model Free Motor Control. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.947-948.