How to Fine-tune the Model: Unified Model Shift and Model Bias Policy Optimization

Hai Zhang, Hang Yu, Junqiao Zhao +5

Designing and deriving effective model-based reinforcement learning (MBRL) algorithms with a performance improvement guarantee is challenging, mainly attributed to the high couplin…