Maximum Entropy Gain Exploration for Long Horizon Multi-goal Reinforcement Learning
This addresses sample efficiency challenges for multi-goal reinforcement learning agents in complex, long-horizon environments, representing a strong incremental improvement over existing methods.
The paper tackles the problem of selecting intrinsic goals for multi-goal reinforcement learning in long-horizon tasks when desired goals are too distant, proposing a strategy that maximizes entropy of achieved goals by focusing on sparsely explored areas. It achieves an order of magnitude better sample efficiency than prior state-of-the-art methods on tasks like maze navigation and block stacking.
What goals should a multi-goal reinforcement learning agent pursue during training in long-horizon tasks? When the desired (test time) goal distribution is too distant to offer a useful learning signal, we argue that the agent should not pursue unobtainable goals. Instead, it should set its own intrinsic goals that maximize the entropy of the historical achieved goal distribution. We propose to optimize this objective by having the agent pursue past achieved goals in sparsely explored areas of the goal space, which focuses exploration on the frontier of the achievable goal set. We show that our strategy achieves an order of magnitude better sample efficiency than the prior state of the art on long-horizon multi-goal tasks including maze navigation and block stacking.