Event-triggered distributed zero-sum differential game for nonlinear multi-agent systems using adaptive dynamic...

Area of Science:

• Control Systems Engineering
• Game Theory
• Artificial Intelligence

Background:

• Distributed multi-agent systems face significant computational and communication burdens.
• Event-triggered control strategies aim to reduce resource consumption by updating controllers only when necessary.
• Zero-sum differential games are crucial for analyzing competitive interactions in dynamic systems.

Purpose of the Study:

• To investigate an event-triggered distributed zero-sum differential game for multi-agent systems.
• To reduce the computational and communication burden in these systems.
• To develop an adaptive control strategy with an event-triggered condition.

Main Methods:

• Derivation of an adaptive event-triggered distributed iterative differential game strategy based on the Minimax principle.
• Approximation of coupled Hamilton-Jacobi-Isaacs (HJI) equations using critic neural networks (NN).
• Design of a persistent of excitation (PE)-free updating law using experience replay and formulation of the system as an impulsive dynamical system.

Main Results:

• Ensured uniformly ultimately bounded (UUB) stability for all closed-loop signals.
• Proved a lower-bounded minimal intersample time, avoiding Zeno behavior.
• Demonstrated significant reduction in controller updates, saving computational and communication resources.

Conclusions:

• The proposed adaptive event-triggered strategy effectively reduces resource demands in distributed multi-agent systems.
• The method guarantees system stability and avoids Zeno phenomena.
• This approach offers a practical solution for resource-constrained multi-agent control problems.