A DIFFERENTIABLE PHYSICS ENGINE FOR DEEP LEARNING IN ROBOTICS
Explore this paper's citation graph
Summary
This paper proposes an implementation of a modern physics engine, which can differentiate control parameters, which is implemented for both CPU and GPU, and shows how such an engine speeds up the optimization process, even for small problems.
- Type
- article
- Published
- 2016-11-05
- Cited by
- 279
- References
- 50
- Access
- Open access
- OpenAlex
- https://openalex.org/W2556096037
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:5763832
Keywords
Artificial intelligence, Robotics, Physics engine, Computer science, Deep learning
References
- The CMA Evolution Strategy: A Comparing Review
- How the body shapes the way we think - a new view on intelligence
- Simulation tools for model-based robotics: Comparison of Bullet, Havok, MuJoCo, ODE and PhysX
- A General Optimization Method using Adjoint Equation for Solving Multidimensional Inverse Heat Conduction
- Resilient Machines Through Continuous Self-Modeling
- Comparing trotting and turning strategies on the quadrupedal oncilla robot
- Nonlinear black-box modeling in system identification: a unified overview
- V-Clip: Fast and Robust Polyhedral Collision Detection
- Approximating polyhedra with spheres for time-critical collision detection
- Long Short-Term Memory
- An aerodynamic optimization method based on the inverse problem adjoint equations
- A Gauss-Seidel like algorithm to solve frictional contact problems
- Automated Design of Complex Dynamic Systems
- Transfer learning of gaits on a quadrupedal robot
- Variational Policy Search via Trajectory Optimization
- NeuroAnimator: fast neural network emulation and control of physics-based models
- On the training of recurrent neural networks
- An implicit time-stepping scheme for rigid body dynamics with Coulomb friction
- Evolving 3D Morphology and Behavior by Competition
- Caffe: Convolutional Architecture for Fast Feature Embedding
Cited by
- Optimization Beyond the Convolution: Generalizing Spatial Relations with End-to-End Metric Learning
- Combining learned and analytical models for predicting action effects from sensory data
- An overview: Paradigm shift of techniques used for educational purposes
- Data-Augmented Contact Model for Rigid Body Simulation
- End-to-End Differentiable Physics for Learning and Control
- ChainQueen: A Real-Time Differentiable Physical Simulator for Soft Robotics
- A Differentiable Augmented Lagrangian Method for Bilevel Nonlinear Optimization
- Combining Physical Simulators and Object-Based Networks for Control
- Physics-as-Inverse-Graphics: Joint Unsupervised Learning of Objects and Physics from Video
- Interactive Differentiable Simulation
- Vid2Param: Online system identification from video for robotics applications
- A Differentiable Programming System to Bridge Machine Learning and Scientific Computing
- Embodied AI beyond Embodied Cognition and Enactivism
- Propagation Networks for Model-Based Control Under Partial Observation
- Differentiable Cloth Simulation for Inverse Problems
- Bayesian Optimization of a Free-Electron Laser.
- Algorithmic differentiation improves the computational efficiency of OpenSim-based trajectory optimization of human movement
- Automatic Differentiation and Continuous Sensitivity Analysis of Rigid Body Dynamics
- Driving Reinforcement Learning with Models
- Using Reinforced Learning Methods to Control Cube Robots
Related papers
- Solving Jacobian Matrix of Parallel Manipulators With Linear Driving Limbs by Using CAD Variation Geometric Approach
- Data-driven Estimation of the Power Flow Jacobian Matrix in High Dimensional Space
- Physical Scaffolding Accelerates the Evolution of Robot Behavior
- Evolutionary Robotics Applied to Hexapod Locomotion: a Comparative Study of Simulation Techniques
- Not all physics simulators can be wrong in the same way