gradSim: Differentiable simulation for system identification and visuomotor control
This addresses the system identification problem for robotics and computer vision by enabling learning from video alone, reducing reliance on costly 3D data, though it builds incrementally on differentiable simulation and rendering techniques.
The paper tackles the problem of estimating physical properties like mass and friction from video without 3D labels, which are often infeasible to obtain, by introducing gradSim, a framework that uses differentiable simulation and rendering to backpropagate from pixels to physical attributes, achieving performance competitive with or better than methods requiring 3D supervision.
We consider the problem of estimating an object's physical properties such as mass, friction, and elasticity directly from video sequences. Such a system identification problem is fundamentally ill-posed due to the loss of information during image formation. Current solutions require precise 3D labels which are labor-intensive to gather, and infeasible to create for many systems such as deformable solids or cloth. We present gradSim, a framework that overcomes the dependence on 3D supervision by leveraging differentiable multiphysics simulation and differentiable rendering to jointly model the evolution of scene dynamics and image formation. This novel combination enables backpropagation from pixels in a video sequence through to the underlying physical attributes that generated them. Moreover, our unified computation graph -- spanning from the dynamics and through the rendering process -- enables learning in challenging visuomotor control tasks, without relying on state-based (3D) supervision, while obtaining performance competitive to or better than techniques that rely on precise 3D labels.