Model predictive game control for personalized and targeted interactive assistance

Area of Science:

• Robotics
• Human-Robot Interaction
• Control Theory

Background:

• Contact robots are vital in physical training and manufacturing.
• Current control methods prioritize system performance over human user intent.
• Optimizing human-robot collaboration requires considering the human's control strategy.

Purpose of the Study:

• To develop a novel controller for contact robots that predicts and incorporates human motor control.
• To investigate human-robot co-adaptation using a model-predictive game (MPG) controller.
• To assess the impact of an assistance meta-parameter on interaction dynamics.

Main Methods:

• Developed a differential game-based controller (MPG controller) to predict human motor control.
• Conducted experiments to analyze human-robot co-adaptation.
• Utilized an assistance meta-parameter to modulate the robot's assistance level.

Main Results:

• MPG controller ensures stable human-robot interaction while reducing human physical effort.
• The robot successfully adapted to individual human interaction behaviors.
• Humans adapted to the robot's strategy, and their behavior was modulated by the assistance meta-parameter.
• Interaction aligned with game theory principles, demonstrating human understanding of the robot's strategy.

Conclusions:

• The MPG controller facilitates optimal human-robot interaction by predicting human intent.
• Human-robot co-adaptation is a key factor in effective collaborative tasks.
• The assistance meta-parameter offers a versatile method for guiding human behavior in robot-assisted applications for training and rehabilitation.