A Differentiable Physics Engine for Deep Learning in Robotics
Top Cited Papers
Open Access
- 7 March 2019
- journal article
- research article
- Published by Frontiers Media SA in Frontiers in Neurorobotics
- Vol. 13, 6
Abstract
An important field in robotics is the optimization of controllers. Currently, robots are often treated as a black box in this optimization process, which is the reason why derivative-free optimization methods such as evolutionary algorithms or reinforcement learning are omnipresent. When gradient-based methods are used, models are kept small or rely on finite difference approximations for the Jacobian. This method quickly grows expensive with increasing numbers of parameters, such as found in deep learning. We propose the implementation of a modern physics engine, which can differentiate control parameters. This engine is implemented for both CPU and GPU. Firstly, this paper shows how such an engine speeds up the optimization process, even for small problems. Furthermore, it explains why this is an alternative approach to deep Q-learning, for using deep learning in robotics. Finally, we argue that this is a big step for deep learning in robotics, as it opens up new possibilities to optimize robots, both in hardware and software.Keywords
All Related Versions
Funding Information
- Agentschap voor Innovatie door Wetenschap en Technologie (131589)
This publication has 21 references indexed in Scilit:
- The CMA Evolution Strategy: A Comparing Review2007
- Resilient Machines Through Continuous Self-ModelingScience, 2006
- An Aerodynamic Optimization Method Based on the Inverse Problem Adjoint EquationsJournal of Computational Physics, 2001
- V-ClipACM Transactions on Graphics, 1998
- A Gauss-Seidel like algorithm to solve frictional contact problemsComputer Methods in Applied Mechanics and Engineering, 1998
- Long Short-Term MemoryNeural Computation, 1997
- Approximating polyhedra with spheres for time-critical collision detectionACM Transactions on Graphics, 1996
- Nonlinear black-box modeling in system identification: a unified overviewAutomatica, 1995
- Evolving 3D Morphology and Behavior by CompetitionArtificial Life, 1994
- A general optimization method using adjoint equation for solving multidimensional inverse heat conductionInternational Journal of Heat and Mass Transfer, 1991