Curiosity Driven Exploration of Learned Disentangled Goal Spaces

TL;DR

This paper demonstrates that using disentangled goal spaces learned via deep representation enhances curiosity-driven exploration in complex environments, enabling more efficient discovery of controllable features and modular goal sampling.

Contribution

It introduces the use of disentangled representations for goal spaces, improving exploration efficiency and enabling modular goal sampling based on learning progress.

Findings

Disentangled goal spaces outperform entangled ones in exploration tasks.

Sampling goals that maximize learning progress enhances exploration.

Learning progress can be used to discover controllable environment features.

Abstract

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.