Objective evaluation of vehicle handling during winter conditions

Sammanfattning: Vehicle handling evaluation is a crucial part of the vehicle development process. The evaluation can be done in two ways, subjectively; by expert test drivers or objectively; by performing repeatable standard manoeuvres usually by steering robots. Subjective testing is resource intensive as prototypes need to be built. Objective testing is less so, as it can be performed in a virtual environment in conjunction with physical testing. In an e˙ort to reduce resources and time used in vehicle development, manufacturers are looking to objective testing to assess vehicle behaviour.Vehicle handling testing in winter strongly relies on subjective testing. This thesis aims to investigate into the usage of objective test strategy to assess vehicle handling behaviour in winter conditions. Manoeuvres and metrics are defined for summer con-ditions, but not for winter. Hence the goal was to define new or modified metrics and manoeuvres custom to winter testing.Data from an objective winter test was obtained and analysed. The manoeuvres used were constant radius (CR), frequency response (FR), sine with dwell (SWD) and throt-tle release in turn (TRIT). The manoeuvres were compared to public standards from the International Organization for Standards (ISO) and National Highway Traÿc Safety Administration (NHTSA) as well as the vehicle manufacturer standards.The data from a reference vehicle is compared to that from three configuration vehicles, one without anti-roll bar in the front, one without rear anti-roll bar and a standard. The di˙erence in the signals between reference and configuration vehicles is compared to the spread in data of the reference vehicle to determine the signal-to-noise ratio in the manoeuvres. The spread of reference data is analysed to determine the distribu-tion and to di˙erentiate between the two test days. To replicate vehicle behaviour in simulation, winter tyre models using brush and Magic Formula model equations were investigated. These were used in a bicycle and a VI-CarRealTime model. The perfor-mance of these are checked and compared. The bicycle model is used in an unscented Kalman filter, to investigate potential improvements in signal processing. The metrics obtained from the study of standards are checked for robustness in winter conditions by analysis of variance (ANOVA) methods. The procedure of selection of metrics from the ANOVA results is explained. Further, the manoeuvres are modified virtually in VI-CarRealTime, from the results of a sensitivity analysis. The di˙erence in metrics between reference and configuration vehicles is maximized.The final results of the thesis were; a test plan with modified manoeuvres and a set of robust metrics. Also containing important information to aid in the execution of the tests. The conclusions drawn were that the noise in winter testing is high, but the di˙erence between vehicles is statistically significant for some robust metrics. The metrics related to yaw rate were in general more robust. Open-loop throttle and steering control in manoeuvres should be avoided as far as possible. A bicycle model is not complex enough to represent vehicle behaviour at high slip angles. Performance increase of a UKF is not justified as to the e˙ort involved.