The Effect of Planning Shape on Dyna-style Planning in High-dimensional State Spaces

TL;DR

This paper investigates how the shape of planning in Dyna-style reinforcement learning affects performance in high-dimensional environments, revealing that longer rollouts improve learning and demonstrating successful use of learned models in complex games.

Contribution

It shows that planning shape significantly influences Dyna's effectiveness and provides the first evidence of learned models successfully used for planning in high-dimensional environments like ALE.

Findings

Longer, fewer rollouts enhance Dyna's performance.

Planning shape impacts the benefit of model-based updates.

Learned models can be effectively used for planning in ALE.

Abstract

Dyna is a fundamental approach to model-based reinforcement learning (MBRL) that interleaves planning, acting, and learning in an online setting. In the most typical application of Dyna, the dynamics model is used to generate one-step transitions from selected start states from the agent's history, which are used to update the agent's value function or policy as if they were real experiences. In this work, one-step Dyna was applied to several games from the Arcade Learning Environment (ALE). We found that the model-based updates offered surprisingly little benefit over simply performing more updates with the agent's existing experience, even when using a perfect model. We hypothesize that to get the most from planning, the model must be used to generate unfamiliar experience. To test this, we experimented with the "shape" of planning in multiple different concrete instantiations of Dyna, performing fewer, longer rollouts, rather than many short rollouts. We found that planning shape has a profound impact on the efficacy of Dyna for both perfect and learned models. In addition to these findings regarding Dyna in general, our results represent, to our knowledge, the first time that a learned dynamics model has been successfully used for planning in the ALE, suggesting that Dyna may be a viable approach to MBRL in the ALE and other high-dimensional problems.