Development and Evaluation of Control Algorithms for the Coupling Process of Autonomous Modular Vehicles

Abstract

The U-Shift vehicle, an innovative Autonomous Modular Vehicle (AMV) concept developed by the German Aerospace Center (DLR), features a modular design that separates the driving module, known as the driveboard, from transport capsules. This design supports various applications and business models, including last-mile delivery and intermodal transport, by enabling capsule exchange through a coupling and decoupling process. However, achieving precise and seamless coupling in confined environments with low-speed maneuvers requires a motion controller capable of handling multiple constraints to ensure accurate path tracking. This thesis focuses on developing and evaluating control algorithms specifically tailored for the U-Shift’s coupling process, with the goal of achieving high accuracy in reverse driving maneuvers and a balanced coordination of longitudinal and lateral motions. After a comprehensive analysis of existing control strategies and vehicle modeling techniques, a Nonlinear Model Predictive Control (NMPC) framework was selected as the foundation for designing a motion controller to meet the requirements of the coupling process. The control framework was implemented and evaluated using MATLAB/Simulink simulation environment, allowing for iterative testing and refinement. Simulations across diverse coupling scenarios demonstrated that the NMPC-based control system consistently achieved accurate trajectory tracking and robust performance, validating its effectiveness for autonomous coupling in modular vehicles. To further enhance the NMPC controller’s performance, Bayesian Optimization was applied to automate the tuning of controller parameters, ensuring robust accuracy across varied conditions. This research contributes a foundational control architecture tailored to the U-Shift’s operational requirements, with promising potential for real-time application in future deployment.