Nonlinear, distributed, and stochastic model predictive control with GRAMPC-(D/S)

Abstract

This article presents the software packages GRAMPC, GRAMPC-D, and GRAMPC-S for nonlinear model predictive control (NMPC). GRAMPC implements a gradient-based augmented Lagrangian method tailored to real-time capability with low computational and memory demands. On this basis, GRAMPC-D provides a modular framework for distributed NMPC of networked systems integrating both optimization and communication, while GRAMPC-S facilitates stochastic NMPC for uncertain systems through different uncertainty propagation methods. After short introductions to the three packages, the respective usage is demonstrated for the example of a nonholonomic mobile robot.