Motion planning for all-terrain wheeled vehicles

Moëz Cherif



Main results:

We address modeling and global motion planning issues for an autonomous wheeled mobile robot moving on an uneven three-dimensional terrain. The basic problem we address is: starting from a given initial configuration of the robot in contact with the terrain, find a feasible trajectory (configurations/velocities and the control torques to be applied on the robot wheels) that moves the robot toward a given desired configuration. The focus in our work is particularly on the issue of dealing with dynamic and wheel/ground interaction constraints. We believe that such an issue is fundamental for characterizing feasible motions in the case of an all-terrain vehicle.
We have developed a suitable multi-level planning algorithm to deal with the intrinsic characteristics of all-terrain locomotion. A key feature of our approach is that it incorporates appropriate physical models to cope with the task dynamics in the motion planning paradigm. The planner is based on a two-level scheme. The high level considers a simplified 2D instance of the motion task and searches a subset of the configuration space of the robot in order to generate nominal sub-goals through which the robot is steered. The local level solves for continuous feasible trajectories and actuator controls to move the robot between neighboring sub-goals in the presence of the entire task constraints. To the best of our knowledge, this is the first implemented planner that solves for feasible trajectories to be performed by a wheeled vehicle on quite complex terrains. Simulation results are presented for the case of a six-wheeled articulated robot as shown in the presented figures.



Recent related publications:




Please, send your comments to Moez.Cherif@inrialpes.fr