# Unsupervised Visuomotor Control through Distributional Planning Networks

arXiv: https://arxiv.org/abs/1902.05542 · HTML: https://mlsanity.com/paper/1902.05542/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 14 Feb 2019.

## Built on by (most cited first, 9 of 9)

- [Guaranteed Discovery of Control-Endogenous Latent States with Multi-Step Inverse Models](https://mlsanity.com/paper/2207.08229/) (2022)
- [Which Mutual-Information Representation Learning Objectives are Sufficient for Control?](https://mlsanity.com/paper/2106.07278/) (2021)
- [Scalable Multi-Task Imitation Learning with Autonomous Improvement](https://mlsanity.com/paper/2003.02636/) (2020)
- [Learning Visually Guided Latent Actions for Assistive Teleoperation](https://mlsanity.com/paper/2105.00580/) (2021)
- [Learning to Optimize in Model Predictive Control](https://mlsanity.com/paper/2212.02603/) (2022)
- [To Follow or not to Follow: Selective Imitation Learning from Observations](https://mlsanity.com/paper/1912.07670/) (2019)
- [Learning Sampling Distributions for Model Predictive Control](https://mlsanity.com/paper/2212.02587/) (2022)
- [A Simple Approach to Continual Learning by Transferring Skill Parameters](https://mlsanity.com/paper/2110.10255/) (2021)
- [Introspective Visuomotor Control: Exploiting Uncertainty in Deep Visuomotor Control for Failure Recovery](https://mlsanity.com/paper/2103.11881/) (2021)

## Abstract (arXiv)

While reinforcement learning (RL) has the potential to enable robots to autonomously acquire a wide range of skills, in practice, RL usually requires manual, per-task engineering of reward functions, especially in real world settings where aspects of the environment needed to compute progress are not directly accessible. To enable robots to autonomously learn skills, we instead consider the problem of reinforcement learning without access to rewards. We aim to learn an unsupervised embedding space under which the robot can measure progress towards a goal for itself. Our approach explicitly optimizes for a metric space under which action sequences that reach a particular state are optimal when the goal is the final state reached. This enables learning effective and control-centric representations that lead to more autonomous reinforcement learning algorithms. Our experiments on three simulated environments and two real-world manipulation problems show that our method can learn effective goal metrics from unlabeled interaction, and use the learned goal metrics for autonomous reinforcement learning.