Event-Driven Receding Horizon Control for Distributed Estimation in Network Systems

TL;DR

This paper introduces an event-driven receding horizon control approach for distributed estimation in network systems, optimizing agent trajectories to minimize estimation error through a novel distributed control framework and machine learning enhancements.

Contribution

It presents a new distributed control method with a novel objective function and machine learning techniques to improve estimation efficiency in multi-agent network systems.

Findings

Significant reduction in estimation error compared to existing methods

Effective distributed control with unimodality property of the objective function

Enhanced computational efficiency via machine learning-based trajectory history exploitation

Abstract

We consider the problem of estimating the states of a distributed network of nodes (targets) through a team of cooperating agents (sensors) persistently visiting the nodes so that an overall measure of estimation error covariance evaluated over a finite period is minimized. We formulate this as a multi-agent persistent monitoring problem where the goal is to control each agent's trajectory defined as a sequence of target visits and the corresponding dwell times spent making observations at each visited target. A distributed on-line agent controller is developed where each agent solves a sequence of receding horizon control problems (RHCPs) in an event-driven manner. A novel objective function is proposed for these RHCPs so as to optimize the effectiveness of this distributed estimation process and its unimodality property is established under some assumptions. Moreover, a machine learning solution is proposed to improve the computational efficiency of this distributed estimation process by exploiting the history of each agent's trajectory. Finally, extensive numerical results are provided indicating significant improvements compared to other state-of-the-art agent controllers.