MPPI-Generic: A CUDA Library for Stochastic Trajectory Optimization

TL;DR

MPPI-Generic is a CUDA-based C++ library that accelerates stochastic trajectory optimization for real-time applications, supporting various control algorithms and customizable models without code modifications.

Contribution

It introduces a flexible, GPU-accelerated library enabling easy integration of multiple stochastic control algorithms with customizable models.

Findings

Achieves real-time performance on various GPUs.

Supports multiple control algorithms within a unified framework.

Allows users to define custom dynamics and cost functions easily.

Abstract

This paper introduces a new C++/CUDA library for GPU-accelerated stochastic optimization called MPPI-Generic. It provides implementations of Model Predictive Path Integral control, Tube-Model Predictive Path Integral Control, and Robust Model Predictive Path Integral Control, and allows for these algorithms to be used across many pre-existing dynamics models and cost functions. Furthermore, researchers can create their own dynamics models or cost functions following our API definitions without needing to change the actual Model Predictive Path Integral Control code. Finally, we compare computational performance to other popular implementations of Model Predictive Path Integral Control over a variety of GPUs to show the real-time capabilities our library can allow for. Library code can be found at: https://acdslab.github.io/mppi-generic-website/ .