A Knowledge-Based Activity Representation for Shared Autonomy Teleoperation of Robotic Arms

Abstract

Physical impairment caused by illness, disability, or age, limits a person’s independence. Performing Activities of Daily Living (ADLs), such as eating and drinking, can be challenging for people with physical impairment since they have no muscle control. Recent studies investigated employing teleoperated robotic arm/ hand systems to facilitate the performance of ADLs by physically impaired users. Robot teleoperation using Brain Computer Interface (BCI) empowers users with the ability to perform ADLs. However, it is difficult to map the signals received from the BCI into commands for controlling a 7 Degrees of Freedom robot arm, therefore a user will find it difficult and uncomfortable to perform such activities using a robot arm controlled with a BCI. To overcome this, this thesis introduces the Constraint-based Activity REpresentation (CARE) framework; a framework for shared autonomy teleoperation of a robotic arm using a BCI. The CARE framework defines activities in an object centric approach, where an object declares a set of symbolic and geometric constraints to be applied while performing different manipulation actions that might involve other objects. The framework assists the user to satisfy these constraints throughout the activity and compensates for the low dimensional signals obtained from the BCI. As a demonstration, the CARE framework was employed to assist an operator in different ADLs, such as pick and place, pouring, door opening, and wiping. The demonstrations were conducted using a simulation of the DLREDANassistance robot controlled using a SpaceMouse.