Distributed multi-agent collision avoidance using robust differential game.

Area of Science:

• Robotics
• Control Theory
• Artificial Intelligence

Background:

• Networked multi-agent systems (MASs) face challenges in collision avoidance and trajectory optimization.
• Existing methods often lack robustness to disturbances and limited observation capabilities.
• Integrating trajectory optimization with collision avoidance in differential games is complex.

Purpose of the Study:

• To develop a novel robust differential game scheme for collision avoidance in MASs.
• To incorporate trajectory optimization objectives alongside obstacle avoidance.
• To address limitations of external disturbances and partial observability.

Main Methods:

• A robust differential game scheme incorporating artificial potential field (APF) for trajectory optimization.
• Distributed robust Hamilton-Jacobi-Isaacs (DR-HJI) equations for local feedback control under limited observation.
• Ant Colony Optimization (ACO) algorithm for determining feedback gains.

Main Results:

• Strategies converge to a local robust Nash equilibrium (R-NE), and a global R-NE under specific conditions.
• Local robust feedback control strategies are constructed without requiring global agent information.
• Simulations demonstrate collision-free navigation, reduced arrival times, and successful target achievement under disturbance.

Conclusions:

• The proposed scheme effectively solves the collision avoidance and trajectory optimization problem for MASs.
• The method is robust to external disturbances and limited observation capabilities.
• The use of DR-HJI equations and ACO provides a practical and efficient solution.