Composing Complex Skills by Learning Transition Policies
Explore this paper's citation graph
Summary
This work proposes a method that can learn transition policies which effectively connect primitive skills to perform sequential tasks without handcrafted rewards, and introduces proximity predictors which induce rewards gauging proximity to suitable initial states for the next skill.
- Type
- article
- Published
- 2018-09-27
- Cited by
- 104
- References
- 39
- OpenAlex
- https://openalex.org/W2909460318
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:89614462
Keywords
Transition (genetics), Computer science, Human–computer interaction, Knowledge management, Chemistry
References
- Temporal credit assignment in reinforcement learning
- Policy Invariance Under Reward Transformations: Theory and Application to Reward Shaping
- Movement templates for learning of hitting and batting
- Learning to select and generalize striking movements in robot table tennis
- Between MDPs and Semi-MDPs: A Framework for Temporal Abstraction in Reinforcement Learning
- MuJoCo: A physics engine for model-based control
- Learning and generalization of motor skills by learning from demonstration
- Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation
- Neural Module Networks
- Unifying Count-Based Exploration and Intrinsic Motivation
- Probabilistic inference for determining options in reinforcement learning
- The Option-Critic Architecture
- Least Squares Generative Adversarial Networks
- FeUdal Networks for Hierarchical Reinforcement Learning
- Curiosity-Driven Exploration by Self-Supervised Prediction
- Count-Based Exploration in Feature Space for Reinforcement Learning
- Emergence of Locomotion Behaviours in Rich Environments
- Learning human behaviors from motion capture by adversarial imitation
- Proximal Policy Optimization Algorithms
- Deep Reinforcement Learning for Dexterous Manipulation with Concept Networks
Cited by
- Combining learned skills and reinforcement learning for robotic manipulations
- Multimodal Model-Agnostic Meta-Learning via Task-Aware Modulation
- Learning to combine primitive skills: A step towards versatile robotic manipulation §
- Scaling simulation-to-real transfer by learning a latent space of robot skills
- IKEA Furniture Assembly Environment for Long-Horizon Complex Manipulation Tasks
- Learning to Coordinate Manipulation Skills via Skill Behavior Diversification
- Program Guided Agent
- To Follow or not to Follow: Selective Imitation Learning from Observations
- Analysis of Optimal Dynamic Manipulation for Robotic Manipulator Based on Pontryagin’s Minimum Principle
- Motion Planner Augmented Action Spaces for Reinforcement Learning
- Accelerating Reinforcement Learning with Learned Skill Priors
- Motion Planner Augmented Reinforcement Learning for Robot Manipulation in Obstructed Environments
- Learning When to Switch: Composing Controllers to Traverse a Sequence of Terrain Artifacts
- Distilling a Hierarchical Policy for Planning and Control via Representation and Reinforcement Learning
- Multimodal interaction for deliberate practice
- Learn Goal-Conditioned Policy with Intrinsic Motivation for Deep Reinforcement Learning
- Learning Setup Policies: Reliable Transition Between Locomotion Behaviours
- Solving Compositional Reinforcement Learning Problems via Task Reduction
- Unsupervised Discovery of Transitional Skills for Deep Reinforcement Learning
- Skill-based Meta-Reinforcement Learning
Related papers
- Proximal Policy Optimization Algorithms
- Modular Multitask Reinforcement Learning with Policy Sketches
- MuJoCo: A physics engine for model-based control
- Between MDPs and Semi-MDPs: A Framework for Temporal Abstraction in Reinforcement Learning
- Methodology of E-Learning and Analysis of Learning Outcomes
- Activity- and taxonomy-based knowledge representation framework