Abstract
The human-robot shared control provides a viable solution in applications such as search and rescue, where human awareness compensates for deficiencies in sensing, perception, and planning. Our focus is a shared system consisting of one human operator and one unmanned ground vehicle (UGV), where the human operator uses a joystick to drive the UGV remotely. The shared control inputs result from the automatic control inputs mixed with the human control inputs based on the mixed-initiative interaction, in which the human operator can manually drive whenever he/she wants. However, the switching between the human operator and the automatic controller may lead to instability of the closed-loop system. Thus, the design of an asymptotically stable shared controller is necessary to maintain the consistent performance of the shared closed-loop system. The shared controller is asymptotically stable when the automatic and human controllers are asymptotically stable. We developed a path-following and trajectory-tracking controller based on the Lyapunov approach, and we introduced a geometric scaling based on the homogeneity approach that continuously rescaled the control inputs to ensure it respected the actuator magnitude limit without affecting the asymptotic stability of the system. Additionally, we developed a collision avoidance controller based on a control barrier function approach, which allows the UGV to maneuver in an environment with static and moving obstacles. Using the passivity-based approach, we design asymptotically stable human control inputs. The automatic controller is mixed with the human controller using a blending law design based on a passive measurement of human intention. We consider two types of UGVs: a vehicle with front-wheel steering and a differential drive robot. A path-following controller using the Ackermann steering model is designed for a four-steering wheel vehicle. The main focus of this work is the differential drive robot system, where we consider static and moving obstacles in the environment. We developed an asymptotically stable shared trajectory tracking with a collision avoidance controller, which was evaluated in a crowded indoor environment using a DS-Automation Sally robot and in an outdoor environment using a Husky-a200. In both evaluations, the human operator uses a force-feedback joystick, which increases human awareness. The developed shared trajectory tracking with collision avoidance controller ensures a smooth transition between the automatic controller and the human; the UGV magnitude limits are respected; the UGV can be deployed in a complex and highly dynamic environment.