Learning Dynamics Model in Reinforcement Learning by Incorporating the Long Term Future
This addresses the challenge of ensuring reliable long-term predictions for efficient planning and exploration in reinforcement learning, representing a novel method for a known bottleneck.
The paper tackles the problem of catastrophic failures in model-based reinforcement learning due to flawed long-term predictions by building a latent-variable autoregressive model that incorporates future information through an auxiliary task, achieving higher reward faster compared to baselines on various tasks and environments.
In model-based reinforcement learning, the agent interleaves between model learning and planning. These two components are inextricably intertwined. If the model is not able to provide sensible long-term prediction, the executed planner would exploit model flaws, which can yield catastrophic failures. This paper focuses on building a model that reasons about the long-term future and demonstrates how to use this for efficient planning and exploration. To this end, we build a latent-variable autoregressive model by leveraging recent ideas in variational inference. We argue that forcing latent variables to carry future information through an auxiliary task substantially improves long-term predictions. Moreover, by planning in the latent space, the planner's solution is ensured to be within regions where the model is valid. An exploration strategy can be devised by searching for unlikely trajectories under the model. Our method achieves higher reward faster compared to baselines on a variety of tasks and environments in both the imitation learning and model-based reinforcement learning settings.