Towards Batch-to-Streaming Deep Reinforcement Learning for Continuous Control
The paper introduces Streaming Soft Actor-Critic (S2AC) and Streaming Deterministic Actor-Critic (SDAC), novel deep reinforcement learning algorithms for continuous control. These methods allow for purely online updates, overcoming hardware limitations of traditional batch RL. They are designed to be compatible with existing batch RL methods, making them suitable for on-device finetuning like Sim2Real transfer.