Differential Game-Based Control for Nonlinear Human-Robot Interaction System With Unknown Desired Trajectory.

Area of Science:

• Robotics
• Control Theory
• Artificial Intelligence

Background:

• Differential games are crucial for human-robot interaction (HRI) trajectory tracking.
• Existing methods are limited to control-affine systems with known trajectories.
• Nonaffine HRI systems and unknown trajectories pose significant challenges.

Purpose of the Study:

• To develop a novel differential game framework for nonaffine HRI systems.
• To enable robots to estimate unknown desired trajectories using Gaussian process regression (GPR).
• To improve trajectory tracking performance in human-robot collaboration.

Main Methods:

• Proposed a differential game framework incorporating a desired trajectory estimator based on GPR.
• Transformed the nonaffine HRI problem into a differentially flat space.
• Derived equilibrium strategies for the transformed problem.
• Proved probabilistic bounds for trajectory tracking error.

Main Results:

• The proposed method outperforms learning-based approaches in robustness, parameter setting, and time efficiency.
• Experimental results show a 55% reduction in tracking error compared to human direct control.
• The trajectory tracking error satisfies a probabilistic bound that tightens with reduced noise variance.

Conclusions:

• The novel differential game framework effectively addresses nonaffine HRI systems with unknown trajectories.
• GPR-based trajectory estimation enhances cooperative control performance.
• The approach offers a robust and efficient solution for human-robot trajectory tracking tasks.