Human-Agent Shared Teleoperation: A Case Study Utilizing Haptic Feedback

Abstract

Even though teleoperation has been widely used in many application areas including nuclear waste handling, underwater manipulation and outer space applications, the required mental workload from human operator still remains high. Some delicate and complex tasks even require multiple operators. Learning from Demonstration (LfD) through teleoperation can provide a solution for repetitive tasks, but in many cases, one task can be a combination of repetitive and varying motion. This paper introduces a shared teleoperation method between human and agent, trained by LfD through teleoperation. In the proposed method, human takes charge of uncertain or critical motion, whereas more mundane and repetitive motion could be carried out through the assistance of the agent. The proposed method has exhibited superior performance as compared to the human-only teleoperation for a peg-in-hole task.