ESTIMATION OF MODEL PARAMETERS FOR RECONSTRUCTING TRAFFIC ACCIDENTS

Even intelligent transportation systems are not always free from traffic accidents. Conversely, the more complex the systems become, the more the methods of analyzing traffic accidents have to be advanced and sophisticated. This paper focuses on how model parameters for reconstructing traffic accidents are estimated analytically. In this study, an impact model and a driving simulation model play important roles. The impact model is based on the two dimensional impact model proposed by Ishikawa. In this model, there are two unknown parameters; normal and tangential restitution coefficients at the impact center. For driving simulation of the pre-impact and the post-impact phases, Sakai's tire model and two-wheel equivalence model were applied for calculating the composition of tire forces at the gravity center. There are four parameters in these driving models: friction coefficient, steering angle, slip ratio of the front tires, and slip ratio of the rear tires. The main purpose here is to establish an optimization method for estimating those parameters. Box's complex algorithm method was applied to minimize an objective function, which is defined as the difference between estimated and observed rest positions of two vehicles. First, to examine the validity of Box's method, artificial accident data was introduced. Then, the model parameters of two actually observed traffic accidents were estimated. It was concluded that the method proposed here was effective for estimating the unknown model parameters.