Dynamic terrain traversal skills using reinforcement learning
Explore this paper's citation graph
Summary
This paper learns controllers that allow simulated characters to traverse terrains with gaps, steps, and walls using highly dynamic gaits using reinforcement learning, with careful attention given to the action representation, non-parametric approximation of both the value function and the policy.
- Type
- article
- Published
- 2015-07-27
- Cited by
- 75
- References
- 44
- OpenAlex
- https://openalex.org/W2014505645
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:17966128
Keywords
Terrain, Traverse, Reinforcement learning, Tree traversal, Computer science
References
- The CMA Evolution Strategy: A Comparing Review
- Beyond keyframing: an algorithmic approach to animation
- Reinforcement Learning and Dynamic Programming Using Function Approximators
- Fast Approximate Nearest Neighbors with Automatic Algorithm Configuration
- A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning
- Data-driven biped control
- Feature-based locomotion controllers
- Responsive characters from motion fragments
- Robust physics-based locomotion using low-dimensional planning
- Motion fields for interactive character locomotion
- Continuation methods for adapting simulated skills
- Falling and landing motion control for character animation
- Iterative Training of Dynamic Skills Inspired by Human Coaching Techniques
- SIMBICON: simple biped locomotion control
- Continuous character control with low-dimensional embeddings
- Learning bicycle stunts
- Optimization-based interactive motion synthesis
- Physically valid statistical models for human motion generation
- Generalized biped walking control
- Precomputing avatar behavior from human motion data
Cited by
- Relationship descriptors for interactive motion adaptation
- Character contact re‐positioning under large environment deformation
- Data‐guided Model Predictive Control Based on Smoothed Contact Dynamics
- Guided Learning of Control Graphs for Physics-Based Characters
- Off-line controller design for reliable walking of ranger
- Terrain-adaptive locomotion skills using deep reinforcement learning
- Unified motion planner for fishes with various swimming styles
- Ballistic motion planning for jumping superheroes
- A Virtual Character Learns to Defend Himself in Sword Fighting Based on Q-Network
- Augmenting sampling based controllers with machine learning
- Learning to Schedule Control Fragments for Physics-Based Characters Using Deep Q-Learning
- The Development, Evaluation and Applications of a Neuromechanical Control Model of Human Locomotion
- DeepLoco
- Exposure
- Recurrent Network-based Deterministic Policy Gradient for Solving Bipedal Walking Challenge on Rugged Terrains
- Emergence of human-comparable balancing behaviours by deep reinforcement learning
- Learning symmetric and low-energy locomotion
- Self-Augmenting Strategy for Reinforcement Learning
- NNWarp: Neural Network-Based Nonlinear Deformation
- DeepMimic
Related papers
- Terrain-based robot navigation using multi-scale traversability indices
- Autonomous Robot Navigation Using Traversability Indices
- Search-based Foot Placement for Quadrupedal Traversal of Challenging Terrain
- Body-terrain interaction affects large bump traversal of insects and legged robots
- Slope traversal controls for planetary exploration rover on sandy terrain
- Reinforcement Learning based Autonomous Vehicle for Exploration and Exploitation of Undiscovered Track