Disputation Ihno Schrot Efficient numerical methods for nonlinear model predictive control with applications in adaptice cruise control

This thesis presents efficient numerical methods for Nonlinear Model Predictive Control (NMPC), with a focus on Ecological Adaptive Cruise Control (EACC) systems for electric vehicles. NMPC is a closed-loop control strategy that allows to respond to disturbances while satisfying system constraints. It does so by using a model of the system dynamics to predict and optimize its future behavior by solving an Optimal Control Problem (OCP) at fixed sampling intervals. EACC extends conventional Adaptive Cruise Control (ACC) by not only maintaining a safe distance to the vehicle ahead but also minimizing energy consumption. To achieve this, EACC leverages predictive information about the upcoming road segment and traffic conditions — making NMPC a natural choice for this application. To efficiently solve the resulting OCPs, we build on the Multi-Level Iteration (MLI) scheme — a highly efficient numerical method for NMPC that is particularly well-suited for control applications like EACC , where computational resources are limited.

Key contributions of this thesis include a novel smooth and shape-preserving multivariate interpolation method that is required to accurately and efficiently treat gridded data used in the modelling process, and new techniques to handle external inputs such as road slope and traffic behavior in the MLI scheme. We further introduce the Sensitivity and External Input Scenario based (SensEIS) feedback strategy, enabling fast online control updates through scenario-based precomputations. We validate our methods through simulations with real-world driving data.