Vehicle dynamics models are important tools for research and development within the automotive industry. The need for such models is steadily rising due to the rapid development of automated driving, as these models allow efficient design and development of the control algorithms and functions. Chalmers Revere Lab supplements research in the field of automated driving and collision avoidance. The lab maintains two test vehicles, a passenger car (Volvo XC90) and a heavy duty truck (6x4 Volvo FH16). The models provided by Volvo are confidential, complicated for implementation as reference models and do not share the same control interface as the test vehicles. The objective of this work is to develop well documented generic and complete open-source models that represent Revere's test vehicles.
The thesis aims to develop three model units per test vehicle. The three model units are a control interface model, a quick simulation model for online prediction and an advanced model for offline simulation prior to track testing. In addition to model development, an investigation has been made on the level of modelling details suitable for automated driving for non-critical highway and city driving on dry asphalt. For instance, a clutch model was implemented to capture the behaviour of starting from a standstill for the test vehicle equipped with an automated manual transmission. The models were developed using Modelica, an object-oriented modelling tool, and Matlab/Simulink. Furthermore, a suitable model architecture was proposed for simulation of automated driving functions.
To assess the validity of the models, the simulation results were compared against experimental data. Data was collected using open-loop test maneuvers and manual driving tests. The simulation results highlight the differences between the simple and advanced model and their accuracy with respect to experimental data. As a use-case, the XC90 advanced model was simulated with a GPS-based autonomous navigation controller. Validation with experimental data showed that the vehicle model is suitable for the development of control algorithms. The simple model is faster than real-time making it suitable for online prediction and the advanced model is real-time capable which was verified with real-time toolbox on Simulink. As a suggestion for future work, the tire model can be improved to handle low-speed parking scenario and a trailer-dolly combination can be added to the tractor model for studies on combination vehicles.