Toward Seamless Transitions Between Shared Control And Supervised Autonomy In Robotic Assistance

Assistive robots aim to help humans with impairments execute motor tasks in everyday household environments. Controlling the end-effector of such robots directly, for instance with a joystick, is often cumbersome. Shared control methods, like Shared Control Templates (SCTs) [1], have therefore been proposed to provide support for robotic control. Moreover, depending on factors such as workload, system trust or engagement, users may like to freely adjust the level of autonomy, for instance by letting the robot complete a task by itself. In this letter, we present a concept for adjustable autonomy in the context of robotic assistance. We extend the SCT approach with an automatic control module that allows the user to switch between Shared Control and Supervised Autonomy at any time during task execution. As both support modes use the same action representation, transitions are seamless. We show the capabilities of this approach in a set of daily living tasks with our wheelchair-mounted robot EDAN and our humanoid robot Rollin' Justin. We highlight how automatic execution benefits from SCT features, like task-related constraints and whole-body control.