Efficient Elitist Cooperative Evolutionary Algorithm for Multi-Objective Reinforcement Learning

Sequential decision-making problems with multiple objectives are known as multi-objective reinforcement learning. In these scenarios, decision-makers require a complete Pareto front that consists of optimal solutions. Such enables to understand the relationship between and make informed decisions from broad range However, existing methods may be unable search for solutions in concave regions or lack global optimization ability, leading incomplete fronts. To address this issue, we propose an efficient elitist cooperative evolutionary algorithm maintains both evolving population elite archive. The archive uses operations various genetic operators guide population, resulting searches experimental results on submarine treasure hunting benchmarks demonstrate effectiveness proposed method solving learning providing set trade-off travel time amount, enabling them flexible based their preferences. Therefore, has potential useful tool implementing real-world applications.