Evolutionary state estimate-based adaptive multi-objective particle swarm optimization

With the widespread application of multi-objective optimization problems (MOPs) in fields such as engineering design, decision analysis, and resource management, traditional multi-objective optimization algorithms face challenges such as a single learning pattern and premature convergence when solving complex problems. To address these issues, this paper proposes an Evolutionary State Estimate (ESE)-based Adaptive Multi-strategy Multi-Objective Particle Swarm Optimization with a Pareto-based Bi-Indicator Infill Sampling Criterion (PBISC), called EAPMPSO. In the proposed approach, dynamic neighborhood and cosine similarity are introduced in the evolutionary state estimate-based multi-strategy particle learning method, along with dynamic oscillating inertia weight, to help offspring particles escape local optima and explore more feasible solutions which means the solutions in the feasible region of a constrained optimization problem. On the ZDT and DTLZ benchmark test suites, EAPMPSO was compared with seven other advanced multi-objective PSO variants such as CMOPSO and multi-objective evolutionary algorithms like AGEMOEAII, achieving remarkable performance when solving functions with different characteristics, ranking first in the results. The results show that the proposed multi-strategy effectively promote the convergence and diversity of multi-objective PSO, providing innovative insights into the development of evolutionary state estimate for multi-objective evolutionary algorithms.