Skills Regularized TaskDecompositionforMulti-task OfflineReinforcementLearning

In the meanwhile, multi-task RL is considered promising to enhance the generality of RL policies and improve the learning efficiency [4, 5, 6, 7, 8, 9, 10].