In this thesis a Model Predictive Control (MPC) based designs will be investigated for online motion control of commercial vehicles.
First the theoretical background will be introduced for the common Dynamic Matrix Control (DMC) and for Quadratic Dynamic Matrix Control (QDMC) which implements linear constraint calculation as a quadratic programming task.
The common two-axle vehicle models will be presented and evaluated for using as reference model for the prediction. Moreover, the paper also discusses a simple vehicle model with one articulation.
The DMC and QDMC will be tested in ideal simulations showing the basic tuning possibilities and controller behaviors. However, it was found that such design is not suitable for online usages due to the large computational load.
With the purpose of optimization, a new design will be presented called Laguerre MPC (LMPC) which is based on QDMC and implements discrete Laguerre networks. Thanks to the Laguerre functions much fewer decision variables and constraints will be required to maintain the same performance quality, this way reducing both code size and turn-around time.
Finally, an additional high-level longitudinal MPC will be created, so the combined controllers can be integrated into a Highway Assist (HWA) function developed by Knorr-Bremse AG. The integrated function will be tested in Software-in-the-loop simulations (SIL) and later as an embedded code on a test vehicle at an enclosed testing site.