Shared Autonomy for Assistive Robotics

Abstract

Assistive robotic arm manipulation has gained increasing popularity with the emergence of lightweight robot arms which employ safe control concepts such as Cartesian impedance control. Controlling such arms using classical tele-operation methods, while easy to implement and user interactive, is prone to inaccuracy and places a lot of mental demand on users. This thesis proposes to mitigate the challenges of task completion and stable grasping faced in reach and grasp tasks while also reducing the cognitive demand and thereby improving the user friendliness of such systems using a shared autonomy approach. The work presented here proposes a shared autonomy scheme using virtual fixturing for trajectory guidance and autonomous hand orientation control towards a stable grasp pose for a 4-DoF user interface. It also develops a method for adapting the graspability map as a tool for online grasp planning for assistive grasping. The pilot user study conducted demonstrated shared autonomy improves task efficiency by reducing the task completion times and trajectory wandering. Participants also expressed a decrease in mental stress while operating the robot arm using the shared autonomy scheme.