Self-Consistent Trajectory Autoencoder: Hierarchical Reinforcement Learning with Trajectory Embeddings

TL;DR

This paper introduces SeCTAR, a hierarchical reinforcement learning model that learns trajectory embeddings through a variational autoencoder framework, enabling effective long-horizon planning and exploration in complex tasks.

Contribution

The paper presents a novel trajectory autoencoder that learns consistent latent-conditioned policies and models, facilitating hierarchical RL with improved planning and exploration capabilities.

Findings

Outperforms standard RL methods on long-horizon tasks

Effective in sparse reward environments

Enables hierarchical reasoning and model-based planning

Abstract

In this work, we take a representation learning perspective on hierarchical reinforcement learning, where the problem of learning lower layers in a hierarchy is transformed into the problem of learning trajectory-level generative models. We show that we can learn continuous latent representations of trajectories, which are effective in solving temporally extended and multi-stage problems. Our proposed model, SeCTAR, draws inspiration from variational autoencoders, and learns latent representations of trajectories. A key component of this method is to learn both a latent-conditioned policy and a latent-conditioned model which are consistent with each other. Given the same latent, the policy generates a trajectory which should match the trajectory predicted by the model. This model provides a built-in prediction mechanism, by predicting the outcome of closed loop policy behavior. We propose a novel algorithm for performing hierarchical RL with this model, combining model-based planning in the learned latent space with an unsupervised exploration objective. We show that our model is effective at reasoning over long horizons with sparse rewards for several simulated tasks, outperforming standard reinforcement learning methods and prior methods for hierarchical reasoning, model-based planning, and exploration.