Leveraging Retrieval-Augmented Embodied Agents to Enhance Robot Manipulation Capabilities

Retrieval-Augmented Embodied Agents (RAEA) leverage an external policy memory bank to enhance the performance of robots in complex manipulation tasks by retrieving relevant policies and integrating them into the learning process.