Adaptive Event-Triggered Bipartite Formation for Multiagent Systems via Reinforcement Learning.

Area of Science:

• Control Theory
• Artificial Intelligence
• Robotics

Background:

• Investigates challenges in nonlinear discrete-time multiagent systems (MASs) including unknown dynamics and antagonistic interactions.
• Addresses the need for energy-efficient control and formation control in complex MAS environments.

Purpose of the Study:

• To develop an online learning and energy-efficient control strategy for MASs.
• To achieve bipartite formation control for MASs with unknown dynamics and antagonistic interactions.
• To design an adaptive event-triggered mechanism for efficient communication and computation.

Main Methods:

• Formulated a distributed combined measurement error function using signed graph theory to convert bipartite formation into a consensus problem.
• Developed an enhanced linearization controller model using dynamic linearization technology.
• Proposed an online learning adaptive event-triggered (ET) actor-critic neural network (AC-NN) framework with optimized NNs and an adaptive ET mechanism.

Main Results:

• Successfully transformed the bipartite formation issue into a consensus issue.
• Demonstrated the convergence of the proposed formation control framework using Lyapunov functions.
• Validated the algorithm's effectiveness through simulation and experimental studies.

Conclusions:

• The proposed online learning adaptive ET AC-NN framework effectively addresses energy-efficient bipartite formation control in nonlinear discrete-time MASs.
• The method is robust to unknown dynamics and antagonistic interactions.
• The approach offers a promising solution for practical applications requiring coordinated multiagent behavior.