Feasibility Checks for Safe Shared Control with Variable Autonomy in Assistive Robotics

Abstract

In robotics, especially on systems that interact with humans, there is an increasing need for adaptable levels of autonomy. These include direct control, shared control, supervised autonomy, and full autonomy. A concrete example of this is the wheelchair-based robot EDAN, where a user may want to initiate the opening of a door in shared control, but then let the system autonomously cross the doorway. It is essential that robot guidance in shared control is robust and helps the human to prevent errors, rather than causing them by guiding the user towards solutions that are ultimately infeasible due to obstacles or limited manipulability. In this letter, we describe a proof-of-concept for fast and iteratively-refined feasibility checks in the context of EDAN. We conduct exploratory experiments in a grasping action with obstacles.