Planning in Dynamic Environments with Conditional Autoregressive Models
This addresses planning challenges in dynamic settings for robotics or AI systems, but appears incremental as it combines existing techniques.
The paper tackles planning in dynamic environments by using conditional autoregressive models over discrete latent spaces with MCTS, introducing a new test environment with varying difficulty and moving elements. The method nearly matches true environment performance, demonstrating its usefulness for model-based planning.
We demonstrate the use of conditional autoregressive generative models (van den Oord et al., 2016a) over a discrete latent space (van den Oord et al., 2017b) for forward planning with MCTS. In order to test this method, we introduce a new environment featuring varying difficulty levels, along with moving goals and obstacles. The combination of high-quality frame generation and classical planning approaches nearly matches true environment performance for our task, demonstrating the usefulness of this method for model-based planning in dynamic environments.