Sample-Efficient and Safe Deep Reinforcement Learning via Reset Deep Ensemble Agents
This work addresses safety and efficiency issues in deep RL for applications requiring reliable performance, though it appears incremental as it builds on existing reset methods.
The paper tackles the problem of primacy bias in deep reinforcement learning, which causes overfitting to early experiences, by proposing a reset-based method using deep ensemble learning to improve sample efficiency and safety, with numerical results demonstrating its effectiveness.
Deep reinforcement learning (RL) has achieved remarkable success in solving complex tasks through its integration with deep neural networks (DNNs) as function approximators. However, the reliance on DNNs has introduced a new challenge called primacy bias, whereby these function approximators tend to prioritize early experiences, leading to overfitting. To mitigate this primacy bias, a reset method has been proposed, which performs periodic resets of a portion or the entirety of a deep RL agent while preserving the replay buffer. However, the use of the reset method can result in performance collapses after executing the reset, which can be detrimental from the perspective of safe RL and regret minimization. In this paper, we propose a new reset-based method that leverages deep ensemble learning to address the limitations of the vanilla reset method and enhance sample efficiency. The proposed method is evaluated through various experiments including those in the domain of safe RL. Numerical results show its effectiveness in high sample efficiency and safety considerations.