Active Perception and Representation for Robotic Manipulation
arXiv:2003.06734
Abstract
The vast majority of visual animals actively control their eyes, heads, and/or bodies to direct their gaze toward different parts of their environment. In contrast, recent applications of reinforcement learning in robotic manipulation employ cameras as passive sensors. These are carefully placed to view a scene from a fixed pose. Active perception allows animals to gather the most relevant information about the world and focus their computational resources where needed. It also enables them to view objects from different distances and viewpoints, providing a rich visual experience from which to learn abstract representations of the environment. Inspired by the primate visual-motor system, we present a framework that leverages the benefits of active perception to accomplish manipulation tasks. Our agent uses viewpoint changes to localize objects, to learn state representations in a self-supervised manner, and to perform goal-directed actions. We apply our model to a simulated grasping task with a 6-DoF action space. Compared to its passive, fixed-camera counterpart, the active model achieves 8% better performance in targeted grasping. Compared to vanilla deep Q-learning algorithms, our model is at least four times more sample-efficient, highlighting the benefits of both active perception and representation learning.
References in corpus (19)
- Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
- Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
- Soft Actor-Critic Algorithms and Applications
- Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling
- Recurrent Models of Visual Attention
- DRAW: A Recurrent Neural Network For Image Generation
- QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation
- Model-Based Reinforcement Learning for Atari
- Hindsight Experience Replay
- Visual Reinforcement Learning with Imagined Goals
- Reinforcement Learning with Unsupervised Auxiliary Tasks
- Visual Foresight: Model-Based Deep Reinforcement Learning for Vision-Based Robotic Control
- Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulation
- Learning to Poke by Poking: Experiential Learning of Intuitive Physics
- Unsupervised Learning of 3D Structure from Images
- Saccader: Improving Accuracy of Hard Attention Models for Vision
- Supersizing Self-supervision: Learning to Grasp from 50K Tries and 700 Robot Hours
- Learning Actionable Representations from Visual Observations
- Simultaneously Learning Vision and Feature-based Control Policies for Real-world Ball-in-a-Cup