PVEs: Position-Velocity Encoders for Unsupervised Learning of Structured State Representations
arXiv:1705.09805
Abstract
We propose position-velocity encoders (PVEs) which learn---without supervision---to encode images to positions and velocities of task-relevant objects. PVEs encode a single image into a low-dimensional position state and compute the velocity state from finite differences in position. In contrast to autoencoders, position-velocity encoders are not trained by image reconstruction, but by making the position-velocity representation consistent with priors about interacting with the physical world. We applied PVEs to several simulated control tasks from pixels and achieved promising preliminary results.
Accepted at Robotics: Science and Systems (RSS 2017) Workshop -- New Frontiers for Deep Learning in Robotics http://juxi.net/workshop/deep-learning-rss-2017/
References in corpus (1)
Cited by in corpus (17)
- Visual Reinforcement Learning with Imagined Goals
- State Representation Learning for Control: An Overview
- A survey on intrinsic motivation in reinforcement learning
- Universal Planning Networks
- SE3-Pose-Nets: Structured Deep Dynamics Models for Visuomotor Planning and Control
- Unsupervised State Representation Learning in Atari
- Dynamics-aware Embeddings
- Learning Actionable Representations from Visual Observations
- Physics-as-Inverse-Graphics: Unsupervised Physical Parameter Estimation from Video
- Low Dimensional State Representation Learning with Reward-shaped Priors
- Uncovering Closed-form Governing Equations of Nonlinear Dynamics from Videos
- Low-Dimensional State and Action Representation Learning with MDP Homomorphism Metrics
- Towards Object Detection from Motion
- Supervise Thyself: Examining Self-Supervised Representations in Interactive Environments
- One-Shot Generation of Near-Optimal Topology through Theory-Driven Machine Learning
- Learn Proportional Derivative Controllable Latent Space from Pixels
- State Representation Learning from Demonstration