Uncertainty analysis of vehicle-pedestrian accident reconstruction based on unscented transformation.

ABSTRACT

In order to investigate the sensitivity of parameters and analyze the uncertainty of reconstructed results in traffic accident, the impact of correlations between parameters on accident reconstruction results was taken into account using uncertainty analysis. Based on unscented transformation (UT), a parameter sensitivity analysis method and an efficient uncertainty analysis method in accident reconstruction were proposed. Sensitivity analysis was performed through the sigma point sets generated by the UT method. A first-order response surface model was constructed to analyze the sensitivity of accident reconstruction parameters combined with regression analysis, which is more flexible and controllable than the general experimental design. For the uncertainty analysis of the reconstructed results, the other methods have been used to demonstrate the validity of the proposed method, including the first second-order method of moments (FOSM), the uncertainty theory, and the Monte Carlo (MC) methods, through analyzing the numerical and real-world cases. The results show that the presented method has high accuracy, significantly reduces the computational burden, and does not depend on the distribution type of variables. When considering the effect of the correlation between parameters of the vehicle-pedestrian crash on accident reconstruction results, the results show that the correlation coefficient between random variables had a much more significant impact on the standard deviation of vehicle speed than on the mean value of vehicle speed. Regardless of negative or positive correlations, the relative error of standard deviation of vehicle speed increased continuously as the correlation increased, reaching 52%. The proposed method is effective and reliable for vehicle collision accident reconstruction under uncertainty and correlation, which can provide more comprehensive information in accident reconstruction.