Shared Control With Efficient Subgoal Identification and Adjustment for Human-Robot Collaborative Tasks

In this article, we describe a novel method to facilitate the process of subgoal identification from a robotics task demonstrated by a human operator. The method defines a unified metric to quantify the human operator's commands, which can be employed to effectively extract subgoal distributions and identify subgoals. We then address the problem of online subgoal adjustment for shared control (SC) applications, which is initiated by the human operator and is necessary to sustain performance in a dynamically changing environment. The adjustment is facilitated by designing a hyperrectangle around the subgoal whose volume is obtained via a hyperbolic slope transition function (HSTF) based on the distance between subgoals. The adjustment actions within the hyperrectangle are encoded by a skill-weighted action integral. A practical implementation of the automatic control input is also provided where proper scaling for input-blending is considered. The developed SC method is corroborated on a scaled hydraulic excavator platform with human operators. Experimental results are presented and discussed.

Automated summary

In this article, a novel method for subgoal identification from a robotics task demonstrated by a human operator is described. A unified metric is defined to quantify the human operator's commands, which helps in extracting subgoal distributions and identifying subgoals effectively. The article also addresses the problem of online subgoal adjustment for shared control applications and provides a practical implementation and experimental results.