Robotic manipulators and exoskeletons have shown great potential in various fields such as motor rehabilitation, physical training, and industrial tasks. Despite their promising applications, their effectiveness has been hindered by the challenge of predicting and adapting to the control strategies employed by human users. This study delves into the notion that treating humans and robots as adaptive agents engaged in interactions guided by game theory principles could enhance their physical collaborations.
The research introduces a novel framework known as model predictive game (MPG) control, rooted in the concepts of differential game theory, to address this challenge. The primary goal of this controller is to predict human motor strategies by evaluating a representative cost function and adjusting robotic assistance accordingly. The validity of this approach was initially verified through simulations and later experimentally tested using a wrist exoskeleton for trajectory tracking tasks.
The outcomes of these studies revealed that MPG control offers several advantages: it decreases human effort while preserving task stability, captures consistent individual motor strategy characteristics over time, and permits adjustments to human motor strategies through an assistance meta-parameter aimed at minimizing human exertion. The findings suggest that humans, in the investigated tasks, comprehend and respond to their partner’s control strategy based on the principles of game theory.
Moreover, the capability of the assistance meta-parameter to steer humans towards specific interaction behaviors paves the way for the development of adaptable robot-assisted systems beneficial for physical training and rehabilitation purposes. This research signifies a significant advancement in understanding and enhancing the collaboration between humans and robotic systems, offering a new avenue for the design and implementation of more effective and versatile robotic technologies for various applications.
About the Author
