Traffic Accident Reconstruction Based on Occupant Trajectories and Trace Identification

Abstract

Momentum-based method and multibody method are often used in accident reconstructions. In this paper, a method that combines momentum-based method and multibody method is presented to analyze a real-world accident with occupant ejection. A multibody model is employed as an effective approach to simulate interaction between vehicle and occupants in the postanalysis of a traffic accident. This model can be used to answer questions that which occupant the driver was, when the accident happened, and the vehicle's speed before occupant ejection. Based on the information of the vehicle turn-over location and vehicle speed at the moment of the accident, some unknown parameters of the vehicle before the accident, such as initial speed, acceleration, and vehicle-ground friction coefficient, can be determined. In order to estimate these unknown parameters, a Kriging surrogate model is first constructed for the vehicle movement simulation. The Markov Chain Monte Carlo (MCMC) simulation is then employed to infer the unknown parameters based on the Kriging surrogate model. The results show that a high degree of consistency has been achieved between the computer simulation model and the accident on-scene information. In addition, the injury of occupants in simulation agrees well with conclusions of autopsy report.