Unsupervised Learning and Exploration of Reachable Outcome Space
This addresses the problem of exploration in sparse-reward environments for reinforcement learning researchers, offering a generalizable solution that is incremental over existing divergent-search methods.
The paper tackles the challenge of reinforcement learning in sparse reward settings by introducing TAXONS, a task-agnostic algorithm that learns diverse policies from high-dimensional observations without task-specific information, resulting in coverage of a significant portion of the ground-truth outcome space.
Performing Reinforcement Learning in sparse rewards settings, with very little prior knowledge, is a challenging problem since there is no signal to properly guide the learning process. In such situations, a good search strategy is fundamental. At the same time, not having to adapt the algorithm to every single problem is very desirable. Here we introduce TAXONS, a Task Agnostic eXploration of Outcome spaces through Novelty and Surprise algorithm. Based on a population-based divergent-search approach, it learns a set of diverse policies directly from high-dimensional observations, without any task-specific information. TAXONS builds a repertoire of policies while training an autoencoder on the high-dimensional observation of the final state of the system to build a low-dimensional outcome space. The learned outcome space, combined with the reconstruction error, is used to drive the search for new policies. Results show that TAXONS can find a diverse set of controllers, covering a good part of the ground-truth outcome space, while having no information about such space.