Hierarchical Imitation Learning with Vector Quantized Models
This addresses the problem of enabling intelligent agents to solve complex, long-horizon decision-making tasks more effectively, representing a novel method for a known bottleneck.
The paper tackled the challenge of learning hierarchical models for multi-level planning from demonstrations in complex tasks, proposing a reinforcement learning approach to identify subgoals and using a vector-quantized generative model for planning, resulting in outperforming state-of-the-art methods on long-horizon problems.
The ability to plan actions on multiple levels of abstraction enables intelligent agents to solve complex tasks effectively. However, learning the models for both low and high-level planning from demonstrations has proven challenging, especially with higher-dimensional inputs. To address this issue, we propose to use reinforcement learning to identify subgoals in expert trajectories by associating the magnitude of the rewards with the predictability of low-level actions given the state and the chosen subgoal. We build a vector-quantized generative model for the identified subgoals to perform subgoal-level planning. In experiments, the algorithm excels at solving complex, long-horizon decision-making problems outperforming state-of-the-art. Because of its ability to plan, our algorithm can find better trajectories than the ones in the training set