Approximate optimal and safe coordination of nonlinear second-order multirobot systems with model uncertainties

Area of Science:

• Robotics
• Control Theory
• Artificial Intelligence

Background:

• Multi-robot systems require sophisticated coordination strategies for complex tasks.
• Ensuring safety, particularly collision avoidance, remains a significant challenge in uncertain environments.
• Existing methods often struggle with nonlinear dynamics and unknown system parameters.

Purpose of the Study:

• To develop an approximate optimal coordination strategy for nonlinear uncertain second-order multi-robot systems.
• To guarantee safety, specifically collision avoidance, during multi-robot operations.
• To address model uncertainties in robot dynamics through adaptive control.

Main Methods:

• Formulation of a collision-free control objective as a coordination optimization problem using novel local error signals.
• Application of approximate dynamic programming (ADP) with critic-only neural networks (NNs) to learn optimal value functions and control policies.
• Redesign of approximated optimal controllers using adaptive laws to compensate for uncertain robot dynamics.

Main Results:

• Demonstrated uniform ultimate boundedness of NN weight estimation errors under specific conditions.
• Achieved safe coordination of multiple robots, effectively handling model uncertainties.
• Validated the controller's effectiveness through numerical simulations.

Conclusions:

• The proposed adaptive control strategy ensures safe and optimal coordination for uncertain multi-robot systems.
• The integration of ADP and NNs provides a robust framework for complex robotic coordination tasks.
• The method offers a promising solution for real-world applications requiring guaranteed collision avoidance.