Reconstructing vehicle dynamics from on-board event data 2019-01-0632
Modern vehicles record dynamic data from a number of on-board sensors for events that could precede a crash. This data can be used to reconstruct the behavior of a vehicle, although the accuracy of these reconstructions has not yet been quantified. Here we evaluated various methods of reconstructing the vehicle dynamics of two 2017 Toyota Corollas based on Vehicle Control History (VCH) data from overlapping events generated by the pre-collision system (PCS), sudden braking, and ABS activation. The vehicles were driven towards a stationary target at 32-64 km/h and then after the pre-collision alarm sounded the vehicle was steered sharply right or left and braked rapidly to rest. VCH data was then imaged for the PCS event at 2 Hz and for the sudden braking and ABS activation events at 6.7 Hz. The steering wheel angle and the vehicle’s longitudinal acceleration, lateral acceleration, and angular rate data were then extracted and used to predict the vehicle position and heading over the duration of the VCH data record preceding the vehicle coming to rest. These predictions were generated by directly integrating the VCH data and by using the VCH data as inputs to PC-Crash simulations. The predicted positions and headings were then compared to the actual position and heading data measured using differential GPS synchronized to the VCH data record. The results of these analyses provide insights into the best methods for reconstructing vehicle dynamics from VCH data and estimates of the errors associated with different reconstruction techniques.
Brandon Tsuge, Mike Yang, Thomas Flynn, Peter Xing, Jonathan Lawrence, Bradley Heinrichs, Gunter P. Siegmund