ACDSLab/MPPI-Generic

What it solves

MPPI-Generic provides a high-performance implementation of Model Predictive Path Integral (MPPI) control, allowing for stochastic trajectory optimization in robotic systems.

How it works

It is a C++/CUDA header-only library that leverages NVIDIA GPUs to parallelize the computation of multiple potential trajectories, enabling the system to find an optimal path based on a cost function.

Who it’s for

Researchers and engineers working on robotics, control theory, and trajectory optimization who need a GPU-accelerated implementation of the MPPI algorithm.

Highlights

• Header-only C++/CUDA library for easy integration.
• Compatible with CUDA 10 and newer (CUDA 11.7+ recommended).
• Built on the Eigen library for linear algebra operations.
• Supports unit testing for verification.