CusADi: A GPU Parallelization Framework for Symbolic Expressions and Optimal Control

Описание к видео CusADi: A GPU Parallelization Framework for Symbolic Expressions and Optimal Control

We present CusADi, a tool to parallelize symbolic CasADi expressions with CUDA for massive GPU parallelization.

With CusADi, we present several applications, such as parallelizing MPC in IsaacGym, computing dynamics algorithms to design RL rewards, and running simulations in parallel to estimate the region of attraction.

Check out our Github: https://github.com/se-hwan/cusadi
And our paper: https://arxiv.org/abs/2408.09662

Комментарии

Информация по комментариям в разработке

Похожие видео