Unsupervised Learning of Goal Spaces for Intrinsically Motivated Goal ExplorationDownload PDF

15 Feb 2018 (modified: 16 Jun 2024)ICLR 2018 Conference Blind SubmissionReaders: Everyone
Abstract: Intrinsically motivated goal exploration algorithms enable machines to discover repertoires of policies that produce a diversity of effects in complex environments. These exploration algorithms have been shown to allow real world robots to acquire skills such as tool use in high-dimensional continuous state and action spaces. However, they have so far assumed that self-generated goals are sampled in a specifically engineered feature space, limiting their autonomy. In this work, we propose an approach using deep representation learning algorithms to learn an adequate goal space. This is a developmental 2-stage approach: first, in a perceptual learning stage, deep learning algorithms use passive raw sensor observations of world changes to learn a corresponding latent space; then goal exploration happens in a second stage by sampling goals in this latent space. We present experiments with a simulated robot arm interacting with an object, and we show that exploration algorithms using such learned representations can closely match, and even sometimes improve, the performance obtained using engineered representations.
TL;DR: We propose a novel Intrinsically Motivated Goal Exploration architecture with unsupervised learning of goal space representations, and evaluate how various implementations enable the discovery of a diversity of policies.
Keywords: exploration; autonomous goal setting; diversity; unsupervised learning; deep neural network
Code: [![github](/images/github_icon.svg) flowersteam/Unsupervised_Goal_Space_Learning](https://github.com/flowersteam/Unsupervised_Goal_Space_Learning)
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 1 code implementation](https://www.catalyzex.com/paper/arxiv:1803.00781/code)
10 Replies