An All-Terrain-Controller for Over-Actuated Wheeled Mobile Robots with Feedforward and Optimization-Based Control Allocation

Mobile robots have successfully been used in rough terrain, most prominently NASA/JPL’s Mars Exploration Rovers. While autonomy and navigation algorithms become more and more capable, the absence of global Navigation satellite systems (GNSS) still poses problems for slip-intensive terrain such as soft sands. Other positioning methods like visual odometry can only be executed in a slow update rate when computation power is limited. This work deals with the problem of executing velocity or acceleration commands from an arbitrary, slowly sampled path planning algorithm in a safer and more robust way. To achieve this goal for a mobile robot with steering capabilities, a model-based All-Terrain-Control (ATC) is proposed. For straight-forward consideration of external forces from the rough terrain and to handle the lack of position and velocity feedback, the proposed controller works on the acceleration level. The over-actuation is solved by an optimization-based control allocation and ultimately, torques and angles are commanded to the wheel drive and steering motors, respectively. It is shown in a realistic co-simulation that the proposed ATC is able to follow a trajectory purely based on the feedback of an inertial measurement unit (IMU). In contrast to a purely geometric reference controller, the ATC mainly calculates the necessary control variables from model knowledge via feedforward control