Contribution à la modélisation et à la commande de robots mobiles à roues

摘要

This dissertation addresses the following problem : given a wheeled mobile robot equip- ped with sensors, considered as perfect, which on-line provide it with its localization. Our aim is 1,0 derive feedback laws which allow it 1,0 move around in an autonomous way. We have first assumed that the robot wheels do not slip on the ground. This is realistic at slow velocities and on grounds with good grip conditions. This assumption imposes that robots cannot move in ail directions. Therefore, they are clearly nonholonomic systems. Their modeling is detailed. It shows that, in case control is the only concern, all wheeled mobile robots can be classified into 5 types. The tracking of a moving reference and the stabilization to a rest configuration are 2 distinct problems for nonholonomic systems. The solutions proposed in the literature may be used to control 4 types of robots. The other one, which gathers robots equipped with several wheels controlled both in orientation and rotation, presents a difficulty Their model encloses singularities. We have proposed, for the tracking, dynamic feedback linearizing laws, and for the stabilization, time-varying feedback laws, which ensure that robots do not meet their singularities. Finally, for the tracking of a moving reference ended by a rest configuration, we have proposed a hybrid scheme combining these laws. Independently, we have also analyzed the convergence rate of time-varying feedback laws. In order 1,0 control wheeled mobile robots at higher velocities and/or on grounds with bad grip conditions, we have, as a second stop, relaxed the non-sliding assumptions. A second model is proposed, relying on the description of the contact type-ground. Since it is undefined when the velocity is zero, we have focused on the tracking of a moving reference. We have proposed static feedback partially linearizing laws, and an adaptive scheme in order that these laws are robust with respect to grip conditions.