Reinforcement learning-based optimal tracking control for uncertain multi-agent systems with uncertain topological...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Robotics

Background:

• Multi-agent systems (MASs) are increasingly applied in intelligent connected vehicles (ICVs) and unmanned aerial vehicles (UAVs).
• Optimal tracking control for MASs with uncertain dynamics and network topologies remains a significant challenge.

Purpose of the Study:

• To develop an observer-based optimal tracking control strategy for uncertain MASs under uncertain network conditions.
• To enhance the robustness and performance of MASs in complex operational environments.

Main Methods:

• An adaptive extended observer using concurrent learning (CL) estimates system states and unknown parameters under relaxed persistence of excitation conditions.
• A Luenberger observer compensates for leader state information under uncertain topologies.
• An actor-critic (AC) neural network (NN) based optimal tracking control algorithm is developed, eliminating the need for state derivative information.

Main Results:

• The concurrent learning observer guarantees convergence of estimated unknown parameters.
• The proposed observers effectively compensate for system uncertainties and network topology variations.
• The actor-critic neural network controller achieves optimal tracking performance without requiring state derivatives.

Conclusions:

• The integrated observer design and reinforcement learning approach provides a robust solution for optimal tracking control in uncertain MASs.
• The developed methodology is validated through numerical simulations, demonstrating its effectiveness for ICVs and UAVs.