Grasp2Vec: Learning Object Representations from Self-Supervised Grasping

Abstract

Well structured visual representations can make robot learning faster and can improve generalization. In this paper, we study how we can acquire effective object-centric representations for robotic manipulation tasks without human labeling by using autonomous robot interaction with the environment. Such representation learning methods can benefit from continuous refinement of the representation as the robot collects more experience, allowing them to scale effectively without human intervention. Our representation learning approach is based on object persistence: when a robot removes an object from a scene, the representation of that scene should change according to the features of the object that was removed. We formulate an arithmetic relationship between feature vectors from this observation, and use it to learn a representation of scenes and objects that can then be used to identify object instances, localize them in the scene, and perform goal-directed grasping tasks where the robot must retrieve commanded objects from a bin. The same grasping procedure can also be used to automatically collect training data for our method, by recording images of scenes, grasping and removing an object, and recording the outcome. Our experiments demonstrate that this self-supervised approach for tasked grasping substantially outperforms direct reinforcement learning from images and prior representation learning methods.

arXiv commentCoRL 2018. Eric Jang and Coline Devin contributed equally to this work

References in corpus (10)

• Distributed Representations of Words and Phrases and their Compositionality
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• End to End Learning for Self-Driving Cars
• QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation
• Hindsight Experience Replay
• Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulation
• Self-Supervised Visual Planning with Temporal Skip Connections
• End-to-End Learning of Semantic Grasping
• The Intentional Unintentional Agent: Learning to Solve Many Continuous Control Tasks Simultaneously
• Multi-Task Domain Adaptation for Deep Learning of Instance Grasping from Simulation

Cited by in corpus (35)

• A Review of Robot Learning for Manipulation: Challenges, Representations, and Algorithms
• High-Fidelity Image Generation With Fewer Labels
• Dex-NeRF: Using a Neural Radiance Field to Grasp Transparent Objects
• S4G: Amodal Single-view Single-Shot SE(3) Grasp Detection in Cluttered Scenes
• Predicting What You Already Know Helps: Provable Self-Supervised Learning
• DGCM-Net: Dense Geometrical Correspondence Matching Network for Incremental Experience-based Robotic Grasping
• Does Unsupervised Architecture Representation Learning Help Neural Architecture Search?
• Data-Efficient Learning for Sim-to-Real Robotic Grasping using Deep Point Cloud Prediction Networks
• Efficient Personalized Speech Enhancement through Self-Supervised Learning
• Uncertainty-Aware Self-Supervised Learning of Spatial Perception Tasks
• Online Self-Supervised Learning for Object Picking: Detecting Optimum Grasping Position using a Metric Learning Approach
• Unsupervised Deep Metric Learning via Auxiliary Rotation Loss
• Self-Supervised GANs via Auxiliary Rotation Loss
• Word2vec to behavior: morphology facilitates the grounding of language in machines
• Proof Artifact Co-training for Theorem Proving with Language Models
• Interferobot: aligning an optical interferometer by a reinforcement learning agent
• Towards a Hypothesis on Visual Transformation based Self-Supervision
• MatchGAN: A Self-Supervised Semi-Supervised Conditional Generative Adversarial Network
• Distribution Estimation to Automate Transformation Policies for Self-Supervision
• Motion2Vec: Semi-Supervised Representation Learning from Surgical Videos
• Self-Supervised Goal-Conditioned Pick and Place
• Object-Agnostic Suction Grasp Affordance Detection in Dense Cluster Using Self-Supervised Learning.docx
• LPaintB: Learning to Paint from Self-Supervision
• Distributed Reinforcement Learning of Targeted Grasping with Active Vision for Mobile Manipulators
• Visual Task Progress Estimation with Appearance Invariant Embeddings for Robot Control and Planning
• Functional Regularization for Representation Learning: A Unified Theoretical Perspective
• Stable deep reinforcement learning method by predicting uncertainty in rewards as a subtask
• A Deep Learning Approach to Grasping the Invisible
• Towards Object Detection from Motion
• Adversarial Feature Training for Generalizable Robotic Visuomotor Control
• GATER: Learning Grasp-Action-Target Embeddings and Relations for Task-Specific Grasping
• A Multi-task Learning Framework for Grasping-Position Detection and Few-Shot Classification
• Zero-shot Policy Learning with Spatial Temporal RewardDecomposition on Contingency-aware Observation
• Adaptation of Quadruped Robot Locomotion with Meta-Learning
• Self-supervised Point Set Local Descriptors for Point Cloud Registration