Modeling and Control of Trust in Human-Robot Collaborative Manufacturing

Abstract

Human-Robot Collaboration (HRC) on the factory floor has opened a new realm of manufacturing in real-world settings. In such applications, a human and robot work together with each other as coworkers while HRC plays a critical role in safety, productivity, and flexibility. In particular, human-robot trust determines his/her acceptance and hence allocation of autonomy to a robot, which alter the overall task efficiency and human workload. Inspired by well-known human factors research, we develop a time-series trust model for human-robot collaboration tasks, which is a function of prior trust, robot performance, and human performance. The robot performance is evaluated by its flexibility to keep pace with the human coworker and is molded as the difference between human and robot speed. The human performance in doing physical tasks is directly related to his/her muscle fatigue level. We use the muscle fatigue and recovery dynamics to capture the fatigue level of the human body when performing repetitive kinesthetic tasks, which are typical types of human motions in manufacturing. The robot speed can be controlled in three different modes: manually by the associate, autonomously through robust intelligence algorithms, or collaboratively by the combination of manual and autonomous inputs. We first simulate a typical 9-h work day for human robot collaborative tasks and implement the proposed trust model and the three control schemes. Furthermore, we experimentally validate our model and control schemes by conducting a series of human-in-the-loop experiments using the Rethink Robotics Baxter robot.