mlsanityData to 14 Oct 2026

Visual Foresight With a Local Dynamics Model

arXiv 2206.14802 ↗First posted 29 Jun 2022cs.RO cs.AI cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Model-free policy learning has been shown to be capable of learning manipulation policies which can solve long-time horizon tasks using single-step manipulation primitives. However, training these policies is a time-consuming process requiring large amounts of data. We propose the Local Dynamics Model (LDM) which efficiently learns the state-transition function for these manipulation primitives. By combining the LDM with model-free policy learning, we can learn policies which can solve complex manipulation tasks using one-step lookahead planning. We show that the LDM is both more sample-efficient and outperforms other model architectures. When combined with planning, we can outperform other model-based and model-free policies on several challenging manipulation tasks in simulation.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 29 of its 41 references that are in the map. First posted on arXiv 29 Jun 2022.

Builds on

29 of 41 references.

References · 29

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  4. 2015Prioritized Experience Replaycited by 1,099
  5. 2018DeepMind Control Suitecited by 956
  6. 2016Pixel Recurrent Neural Networkscited by 816
  7. 2018Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Modelscited by 774
  8. 2018QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulationcited by 743
  9. 2019Model-Based Reinforcement Learning for Ataricited by 502
  10. 2016Learning Hand-Eye Coordination for Robotic Grasping with Deep Learning and Large-Scale Data Collectioncited by 316
  11. 2016Deep Visual Foresight for Planning Robot Motioncited by 285
  12. 2016Continuous Deep Q-Learning with Model-based Accelerationcited by 157
17 more
  1. 2016Learning to Poke by Poking: Experiential Learning of Intuitive Physicscited by 154
  2. 2019TossingBot: Learning to Throw Arbitrary Objects with Residual Physicscited by 137
  3. 2018Learning Synergies between Pushing and Grasping with Self-supervised Deep Reinforcement Learningcited by 128
  4. 2019Unsupervised Learning of Object Structure and Dynamics from Videoscited by 92
  5. 2021Equivariant Q Learning in Spatial Action Spacescited by 62
  6. 2019Structured Object-Aware Physics Prediction for Video Modeling and Planningcited by 43
  7. 2020VisuoSpatial Foresight for Multi-Step, Multi-Task Fabric Manipulationcited by 40
  8. 2020Spatial Action Maps for Mobile Manipulationcited by 31
  9. 2017Deep Q-learning from Demonstrationscited by 28
  10. 2017End-to-End Learning of Semantic Graspingcited by 20
  11. 2020Policy learning in SE(3) action spacescited by 14
  12. 2019Robot Learning of Shifting Objects for Grasping in Cluttered Environmentscited by 13
  13. 2018Visual Robot Task Planningcited by 11
  14. 2022Transporters with Visual Foresight for Solving Unseen Rearrangement Taskscited by 9
  15. 2018Deictic Image Maps: An Abstraction For Learning Pose Invariant Manipulation Policiescited by 3
  16. 2019Learning Visual Affordances with Target-Orientated Deep Q-Network to Grasp Objects by Harnessing Environmental Fixturescited by 3
  17. 2021Learning a Generative Transition Model for Uncertainty-Aware Robotic Manipulationcited by 1

Built on by

Not cited yet.