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

TL;DR

CusADi is a GPU-based extension of CasADi that enables large-scale parallelization of symbolic expressions and optimal control problems, significantly accelerating controller training and simulation tasks.

Contribution

It introduces CusADi, a novel GPU parallelization framework for symbolic expressions and optimal control, facilitating efficient large-scale computations.

Findings

Ten-fold speedup over CPU-based MPC implementations

Effective parallel simulation and parameter sweeps

Enhanced policy training capabilities

Abstract

The parallelism afforded by GPUs presents significant advantages in training controllers through reinforcement learning (RL). However, integrating model-based optimization into this process remains challenging due to the complexity of formulating and solving optimization problems across thousands of instances. In this work, we present CusADi, an extension of the CasADi symbolic framework to support the parallelization of arbitrary closed-form expressions on GPUs with CUDA. We also formulate a closed-form approximation for solving general optimal control problems, enabling large-scale parallelization and evaluation of MPC controllers. Our results show a ten-fold speedup relative to similar MPC implementation on the CPU, and we demonstrate the use of CusADi for various applications, including parallel simulation, parameter sweeps, and policy training.