Auto-conditioned Recurrent Mixture Density Networks for Learning Generalizable Robot Skills
Explore this paper's citation graph
Summary
This paper introduces a state transition model (STM) that generates joint-space trajectories by imitating motions from expert behavior and shows in real robot experiments that the learned STM can quickly generalize to unseen tasks and synthesize motions having longer time horizons than the expert trajectories.
- Type
- preprint
- Published
- 2018-09-29
- Cited by
- 11
- References
- 42
- Access
- Open access
- OpenAlex
- https://openalex.org/W2931117926
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:84186967
Keywords
Computer science, Trajectory, Robot, Context (archaeology), Planner
References
- Sampling-based algorithms for optimal motion planning
- Generating Sequences With Recurrent Neural Networks
- Batch Informed Trees (BIT*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs
- A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning
- A survey of robot learning from demonstration
- Apprenticeship learning via inverse reinforcement learning
- Gaussian Processes for Data-Efficient Learning in Robotics and Control
- Bullet physics simulation
- Robot learning from demonstration
- Long Short-Term Memory
- CHOMP: Gradient optimization techniques for efficient motion planning
- Better Generative Models for Sequential Data Problems: Bidirectional Recurrent Mixture Density Networks
- Particle Filters in Robotics
- Speech recognition with deep recurrent neural networks
- Human-level control through deep reinforcement learning
- Design and use paradigms for Gazebo, an open-source multi-robot simulator
- Improving Multi-Step Prediction of Learned Time Series Models
- Learning Manipulation Trajectories Using Recurrent Neural Networks
- Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model
- (CAD)^2RL: Real Single-Image Flight without a Single Real Image
Cited by
- Attention-based Sampling Distribution for Motion Planning in Autonomous Driving
- Learning a Decentralized Multi-arm Motion Planner
- Cost-to-Go Function Generating Networks for High Dimensional Motion Planning
- Projection: a mechanism for human-like reasoning in Artificial Intelligence
- MoVEInt: Mixture of Variational Experts for Learning Human–Robot Interactions From Demonstrations
- Probabilistic Sampling Networks for Hybrid Structure Planning in Semi-Structured Environments
- Sub-Goal Trees - a Framework for Goal-Based Reinforcement Learning
- Zero-Shot Imitating Collaborative Manipulation Plans from YouTube Cooking Videos
- MoVEInt: Mixture of Variational Experts for Learning HRI from Demonstrations
- Auto-conditioned Recurrent Mixture Density Networks for Learning Generalizable Robotic Manipulation Skills
- Learning Collaborative Action Plans from YouTube Videos
Related papers
- CHEF: A Model of Case-Based Planning
- The Professional Urban Planner
- Dynamic instructional planning for an intelligent physiology tutoring system
- 1P1-B02 Trajectory Tracking Control and Collision Avoidance Trajectory of Rotary Crane
- Representing And Recognizing Motion Trajectories: A Tube And Droplet Approach
- Trajectory Synthesis and Inverse Dynamics Formulation for Minimal Vibrational Excitation for Flexible Structures based on Trajectory Patterns