Curiosity Driven Exploration of Learned Disentangled Goal Spaces
This work addresses the challenge of autonomous exploration for robots in complex environments, representing an incremental improvement over prior methods that relied on engineered or simple learned goal spaces.
The paper tackled the problem of efficient exploration in complex environments with multiple objects or distractors by using a disentangled goal space, showing that it leads to better exploration performance than an entangled goal space and enables modular goal sampling to maximize learning progress.
Intrinsically motivated goal exploration processes enable agents to autonomously sample goals to explore efficiently complex environments with high-dimensional continuous actions. They have been applied successfully to real world robots to discover repertoires of policies producing a wide diversity of effects. Often these algorithms relied on engineered goal spaces but it was recently shown that one can use deep representation learning algorithms to learn an adequate goal space in simple environments. However, in the case of more complex environments containing multiple objects or distractors, an efficient exploration requires that the structure of the goal space reflects the one of the environment. In this paper we show that using a disentangled goal space leads to better exploration performances than an entangled goal space. We further show that when the representation is disentangled, one can leverage it by sampling goals that maximize learning progress in a modular manner. Finally, we show that the measure of learning progress, used to drive curiosity-driven exploration, can be used simultaneously to discover abstract independently controllable features of the environment.