Model Predictive Control for Mobile Robots with Actively Reconfigurable Chassis

Abstract

Actively reconfigurable chassis enable planetary mobile robots to access more varieties of terrain. While typical approaches for exploiting such mechanisms reply on feedback control, it is beneficial to consider actively controlled elements at planning time rather than during motion execution. In this paper we present an approach for extending work in model-predictive trajectory generation to actively reconfigurable chassis. The motion planner uses a kinematic motion model and a terrain shape model to determine sequences of actions that minimize a cost function over vehicle attitude by modifying the shape of the velocity, curvature, and chassis configuration profiles. Simulation and field results are presented demonstrating the benefits of this technique on a prototype mobile robot for lunar excavation.