Asynchronous self-triggered nash learning in distributed multiplayer systems

Yang, Gaofu

doi:10.1016/j.neucom.2026.134446

Asynchronous self-triggered nash learning in distributed multiplayer systems

2026-07-01

1

PRE

AI

Y

Yang, Gaofu

宋睿卓 (Ruizhuo Song) *

DOI：10.1016/j.neucom.2026.134446

Original

Original request for help

Save

Abstract

En 中文

Sparse communication networks hinder direct information exchange among players, rendering conventional ob server designs based on synchronous updates ineffective and making fully connected network-based approaches impractical. To address this issue, this article proposes an off-policy model-free reinforcement learning algo rithm based on an asynchronously triggered observer to solve for the Nash equilibrium in networked multiplayer systems. Under a constrained communication topology, where players cannot directly access non-adjacent in formation, a self-triggered observer is designed to estimate the required inaccessible information. Furthermore, an asynchronous learning triggering condition is constructed based on the maximum inter-event interval of the observer and the Hamilton-Jacobi-Bellman residual, enabling each player to update its control policy indepen dently. A theoretical analysis within a Banach space framework establishes the convergence of the proposed algorithm, and simulation studies validate its effectiveness.

Keywords:

Adaptive dynamic programming

Event-triggered

Non-zero-sum game

Asynchronous policy iteration

Asynchronous self-triggered nash learning in distributed multiplayer systems

Abstract

Journal

Organization

Cited Papers