Blending of human and obstacle avoidance control for a high speed mobile robot

Abstract

Humans remain in the loop in teleoperation because they have some knowledge that the robot they are controlling does not. At the same time teleoperated robots can be programmed to be very good at many tasks, such as avoiding obstacles. Therefore, sharing control between a human and semi-autonomous behaviors on a robot has great potential. This paper presents a model predictive control (MPC) shared control framework for blending human inputs with autonomous behavior inputs. This work adds consideration of how the human input differs from that of an autonomous controller in addition to threat of collision. The framework is applied to a high speed differential drive robot moving through an obstacle field. Preliminary tests by the authors compared the MPC shared control framework to switching obstacle avoidance on/off and the proposed MPC shared control gives the human up to 26% more control with a 35% reduction in collision penalty. Compared to pure human control, MPC shared control demonstrated a 66% reduction in collision penalty. Results show promise for increased user control with better performance.