Efficient In-Vehicle Calibration by the Usage of Automation and Enhanced Online DoE Approaches

Data-driven modeling techniques in combination with efficient test planning procedures are widely used for calibration tasks. Such modeling techniques will also be required for a successful transfer of vehicle calibration activities to the test bench and finally to the office in a so called “Road to rig to desktop” approach. A reduction in development time and cost for prototypes are the major drivers behind this development. However, specific aspects of vehicle calibration like drivability and final approval will still be conducted mainly in the vehicle for the next years. Well established approaches and structured processes are carried over from test bench calibration to the vehicle to ensure that all aspects of vehicle calibration can be managed within an appropriate time and cost frame. An effective drivability calibration requires the objectification of the subjective impressions, but also the appropriate modeling and optimization of the vehicle behavior. Gaussian Process models and their extensions are applied to this task. The typically manual iteration loops are replaced by an automated procedure, containing: online test planning, maneuver driving, evaluation, and modeling. The approach is demonstrated on selected use cases of the drivability calibration domain.