A collaboration-based particle swarm optimizer with history-guided estimation for optimization in dynamic environments. (15th April 2019)

• Record Type: Journal Article
• Authors: Cao, Leilei; Xu, Lihong; Goodman, Erik D.
• Is Part Of: Expert systems with applications. Volume 120(2019)
• Page Start: 1
• Page End: 13
• Publication Date: 2019-04-15
• Subjects: Intelligent systems -- Dynamic optimization -- Particle swarm optimizer -- Collaborative mechanism -- History-guided estimation; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2018.11.020
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: A collaborative mechanism is proposed to improve particle swarm optimization. A worst-replacement scheme is proposed to update particles' positions. The trajectory of the best particle during optimizing is stored to an external archive. The stored solutions estimate a promising optimal region in a new environment. The proposed algorithm shows a competitive power in dynamic optimization problems. Abstract: Optimization problems widely exist in many expert and intelligent systems, e.g., greenhouse intelligent control systems in agriculture, energy management systems for hybrid electric vehicle, and job shop scheduling systems in manufacture. For the optimization problems in these systems, the objective functions may change over time. This kind of problem is usually called as dynamic optimization problems (DOPs) or optimizing in dynamic environments. The optimization algorithm plays an important role in designing an expert and intelligent system. In this paper, we present a novel particle swarm optimizer for optimization in dynamic environments. We introduce two schemes to improve performance of particle swarm optimization in dynamic environments. Firstly, the classical particle swarm optimization is enhanced by a collaborative mechanism, in which a target particle learns from another randomly selected particle and the global best one in the swarm. Instead of moving to the new position directly, a worst replacement operator is used to update the swarm, whereby the worst particle in the swarm moves to the better newly generated position. During optimizing, the best solution in each generation is stored. When an environmental change is detected, the historical solutions are retrieved to collaborate with some newly generated solutions to adapt to the new environment. The performance of the proposed algorithm is compared with several reported algorithms over the benchmark problems. Experimental results indicate that the proposed algorithm offers superior performance compared with the competitors.