Self-adaptive multi-objective differential evolution algorithm with first front elitism for optimizing network usage in networked control systems

Abstract

In networked control systems, by means of event-triggered transmission, it is possible to reduce the network usage, keeping the control system performance at satisfactory levels. There are several schemes for event-triggered transmission. In this study, we propose a multi-objective optimization problem to tune the event-triggered mechanisms. On solving the proposed problem by means of multi-objective evolutionary optimization, a set of efficient solutions is generated with different tradeoffs between control system performance and the number of transmissions. To solve the proposed problem, we also developed an improved multi-objective differential evolution algorithm that includes a self-adaptive mechanism, dynamic crowding distance operator, and novel elitism of the first front. The proposed method is applied to tune decentralized event-triggered mechanisms for a controller given a priori, considering random network-induced delays and packet loss. Two case studies are present ed, comparing the performance of eight different decentralized event-triggered schemes, analyzing the selection of the sampling period, and demonstrating the efficacy of the proposed tuning method.