Fully Data-Driven Robust Output Formation Tracking Control for Heterogeneous Multiagent System With Multiple Leaders...

Area of Science:

• Control Systems
• Artificial Intelligence
• Robotics

Background:

• Traditional multiagent system (MAS) control often requires extensive prior knowledge.
• Actuator faults in MAS can degrade performance and compromise stability.
• Achieving robust formation tracking under uncertainty is a significant challenge.

Purpose of the Study:

• To develop a fully data-driven control framework for robust output formation tracking in MAS with actuator faults.
• To eliminate the need for detailed system prior knowledge in the control design.
• To ensure reliable formation control despite system uncertainties and faults.

Main Methods:

• A hierarchical, three-stage learning and control framework utilizing online data.
• Design of a distributed adaptive observer for leader state coordination and dynamics estimation.
• Application of an off-policy reinforcement learning (RL) method for acquiring feedback gains from partial observations.
• Integration of adaptive neural networks and robust compensations for fault-tolerant control.

Main Results:

• Successful estimation of unknown system dynamics and implicit acquisition of output regulation solutions.
• Effective learning of essential system models through the RL process.
• Robust output formation tracking achieved under actuator faults (efficiency loss, bias) using a model-free approach.
• Demonstrated stability of the learning and control algorithms through analysis.

Conclusions:

• The proposed fully data-driven method effectively addresses the robust output formation tracking problem in MAS with actuator faults.
• The hierarchical framework, leveraging adaptive observers and RL, offers a practical solution without requiring extensive system knowledge.
• Simulation results validate the efficacy and robustness of the developed control strategy.