DiffCloth: Differentiable Cloth Simulation with Dry Frictional Contact

Designing this twirl dress with our differentiable cloth simulator DiffCloth. The goal is to optimize the material parameters so that the apex angle of the cone it forms after a twirl reaches a desired value. Left and middle: the optimized dress before and after a twirl. Top right and bottom right: motion sequences of the twirl dress before and after material parameter optimization using our differentiable simulator.
This work presents a differentiable cloth simulator that extends a state-of-the-art cloth simulator based on Projective Dynamics (PD) and with dry frictional contact. A novel back-propagation technique is proposed to accelerate gradient computation. We demonstrate the effectiveness of the simulator through inverse tasks including system identification, trajectory optimization for assisted dressing, closed-loop control, inverse design.