Abstract: This paper introduces a self-guided hierarchical reinforcement learning (SG-HRL) framework for complex long-horizon dual-arm manipulation tasks. Our approach addresses three fundamental ...