Distributed Safe Event-Triggered Optimal Data-Driven Learning for Nonlinear MASs

doi:10.1109/jiot.2026.3742493

Distributed Safe Event-Triggered Optimal Data-Driven Learning for Nonlinear MASs

2026-10-12

0

PRE

AI

DOI：10.1109/jiot.2026.3742493

Original

Original request for help

Save

Abstract

En 中文

This paper addresses the problem of distributed safe dynamic event-triggered (ET) optimal control for nonlinear multi-agent systems (MASs) under differential graphical games. The safe decision-making problem is formulated as the solution of coupled Hamilton-Jacobi-Bellman (HJB) equations. First, a distributed observer is designed to estimate the leader’s state for each agent using only local neighborhood information, thereby eliminating dependence on the global topology structure. Furthermore, distinct from existing dynamic ET methods, a safe-zone-based fully dynamic ET mechanism is developed by incorporating a smooth scheduling function to enhance resource utilization and tracking performance under obstacle-avoidance constraints. To relax the requirement for system dynamics, integral reinforcement learning (IRL) is used via integral operations along system trajectories. Consequently, an actor-critic neural network (NN) framework, embedded with a state-following kernel (SFK) function, is established to approximate cost function while avoiding noise-susceptible measurements for safety-critical systems. Finally, simulation results are presented to validate the effectiveness and validity of the proposed control scheme.

Keywords:

Nonlinear multi-agent systems

event-triggered control

neural networks

optimal control

obstacle avoidance

Distributed Safe Event-Triggered Optimal Data-Driven Learning for Nonlinear MASs

Abstract

Journal

Organization

Cited Papers