The code in this repository is based on the manuscript “A Decomposition Optimization-Based Multi-objective Reinforcement Learning Algorithm for Obtaining Non-convex Pareto Fronts”, which is submitted to IEEE Transactions on Neural Networks and Learning Systems

This is the demo results of traninng process for the proposed MORL/D-VR algorithm, where the first column represents the timesteps and the other is about the metrics.

The following is about learning curves of the average obtained by MORL/D-VR and MORL/D on DST-1, DST-2, DST-3, and DST-4 tasks: