Observer-based event-triggered optimal control for unknown nonlinear stochastic multi-agent systems with input constraints

This paper addresses the observer-based event-triggered optimal control (ETOC) for unknown non-linear Ito-type stochastic multi-agent systems (SMASs) with input constraints. To begin with, the event -triggered stochastic Hamilton-Jacobi-Bellman (HJB) equation with input constraints is presented, and a sufficient criterion on optimal mean-square leader-following consensus of constrained-input SMASs is derived. Next, a novel event-triggered policy iteration algorithm of constrained-input SMASs is de-signed to obtain the ETOC strategy. Then, an identifier-critic framework is designed where the observer -based identifier network is utilized to recover the knowledge of unknown stochastic dynamics and the constrained-input approximate event-triggered optimal controller is designed via event-triggered adaptive critic designs (ET-ACDs). Moreover, it is proved that the Zeno behavior can be excluded in the sense of expectation. Finally, we present two examples to further verify the validity of the ETOC scheme. & COPY; 2023 The Franklin Institute. Published by Elsevier Inc. All rights reserved.

Automated summary

This paper focuses on the observer-based event-triggered optimal control (ETOC) for unknown non-linear Ito-type stochastic multi-agent systems (SMASs) with input constraints. It presents the event-triggered stochastic Hamilton-Jacobi-Bellman (HJB) equation with input constraints and derives a sufficient criterion for optimal mean-square leader-following consensus of constrained-input SMASs. A novel event-triggered policy iteration algorithm is designed to obtain the ETOC strategy for constrained-input SMASs. An identifier-critic framework is designed using an observer-based identifier network to recover the knowledge of unknown stochastic dynamics and an event-triggered approximate optimal controller is designed using event-triggered adaptive critic designs (ET-ACDs). It is also proven that Zeno behavior can be excluded in terms of expectation. Two examples are provided to further verify the validity of the ETOC scheme.