Optimized predefined-time control for high-order nonlinear MASs via ICA and reinforcement learning

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Nonlinear Dynamics

Background:

• Multi-agent systems (MASs) often face challenges in achieving coordinated behavior due to their high-order nonlinear dynamics.
• Existing consensus algorithms may lack adaptability and precise convergence time guarantees.
• Predefined-time control offers faster convergence than traditional finite-time methods but requires careful design for practical implementation.

Purpose of the Study:

• To develop an adaptive optimized controller for high-order nonlinear MASs enabling practical predefined-time consensus.
• To ensure tracking errors are bounded within prescribed performance limits.
• To achieve convergence to a compact residual set within an adjustable predefined time, independent of initial conditions.

Main Methods:

• Integration of reinforcement learning (RL) and sliding mode control (SMC) within an identifier-critic-actor (ICA) framework.
• Utilization of prescribed-performance control to shape tracking errors.
• Application of a projection-correction technique in neural-network (NN) weight updates to prevent drift and ensure stability.
• Minimization of a cost function balancing consensus error and control input.

Main Results:

• The proposed controller successfully achieves practical predefined-time consensus in high-order nonlinear MASs.
• Convergence time is adjustable via design parameters and independent of system initial states.
• The projection-correction technique effectively stabilizes neural network weights.
• Numerical simulations confirm the controller's effectiveness and performance.

Conclusions:

• The adaptive ICA framework with RL and SMC provides an effective solution for predefined-time consensus in complex MASs.
• The method offers enhanced control over convergence time and stability, crucial for real-world applications.
• This approach advances the state-of-the-art in coordinated control for nonlinear multi-agent systems.