A High Performance Multi-objective Evolutionary Algorithm Based on the Principles of Thermodynamics

In this paper, we propose a high performance multi-objective evolutionary algorithm (HPMOEA) based on the principles of the minimal free energy in thermodynamics. The main innovations of HPMOEA are : (1) providing of a new fitness assignment strategy by combining Pareto dominance relation and Gibbs entropy, (2) the provision of a new criterion for selection of new individuals to maintain the diversity of the population. We use convergence and diversity to measure the performance of the proposed HPMOEA, and compare it with the other four well-known multi-objective evolutionary algorithms (MOEAs): NSGA II, SPEA, PAES, TDGA for a number of test problems. Simulation results show that the HPMOEA is able to find much better spread of solutions and has better convergence near the true Pareto-optimal front on most problems.