Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience
Evo-RL is an iterative reinforcement learning method that combines human interventions with advantage-guided experience selection to iteratively refine the policy on a real-world high-precision manipulation task.