Title of Presentation: Terrain Aware Inversion of Predictive Models for Planetary Rovers

Primary (Corresponding) Author: Alonzo Kelly

Organization of Primary Author:  Carnegie Mellon University


Abstract:  Planetary rovers will continue to evolve in the direction of enhanced mobility and more challenging terrain. Once the mechatronics are in place, it becomes the job of software to exploit the mobility of the platform to the maximum degree possible while minimizing exposure to mobility risk.

Local motion planning is a common component in autonomous mobility architectures which is responsible for using relatively high fidelity models of vehicle motion to correctly predict the consequences of candidate actions. More correct predictions leads to more intelligent behavior, more effective science, and reduced mobility risk.

We favor an approach to precision motion control that uses parameterization to enable efficient search of the set of all feasible candidate motions. Our formulation also enables an efficient inversion of the equations of motion to compute the precise controls necessary to achieve a desired position and orientation while following the contours of the terrain under arbitrary wheel terrain interactions. All higher level rover behaviors can benefit from such precision, terrain aware controls. Applications to instrument placement, wheel slip compensation, obstacle avoidance, and regional mobility planning will be presented.