An intelligent method for accident reconstruction involving car and e-bike coupling automatic simulation and multi-objective optimizations

Car-electric bicycle (e-bike) accidents have been the subject of strong attention due to the widespread usage of e-bikes and a high casualty rate for their riders. Manually conducted accident reconstruction is based on the trial-and-error method with a limited number of parameter combinations, which makes it time-consuming and subjective. This paper aims to develop an intelligent method for accurate, high-efficient reconstruction of accidents involving cars and e-bikes. First, an automatic operation framework, which can drive the MADYMO program and perform results analysis automatically, was built with four multi-objective optimization algorithms available - NSGA-Ⅱ, NCGA, AMGA, and MOPS; The optimization condition was controlled with 12 design variables, 5 objective functions, and 3 constraints. Then, a real e-bike accident with surveillance video was reconstructed through the proposed framework to verify its validity using comparisons of simulated and actual rest positions, initial variables, kinematic response, and head injury. Lastly, the simulation data were used to study the effects of the initial variables on objectives with a multiple linear regression model. The results showed that it took only about 24 h in total for optimization with 480 automatic operations. Optimal conditions were searched at run numbers of 469, 430, 323, and 474 for NSGA-Ⅱ, NCGA, AMGA, and MOPS, respectively. NSGA-Ⅱ had the best performance for e-bike accident reconstruction with average errors of objectives below 5%; Good consistencies for the rider's kinematic response in three stages after collision were observed between simulations and screenshots from the surveillance video, as well as for velocities between the simulation and those estimated from the surveillance video and for head injury between the simulation and the medical report. In contrast to the subjective trial-and-error method that highly depends on the analyst's intuition and experience, this intelligent method is based on multi-objective optimization theory, with which results can be optimized in terms of the automatic change of initial variables. All the above comparisons demonstrate that the method is valid for effectively improving efficiency without simultaneously compromising accuracy. This intelligent method, coupling automatic simulation and multi-objective optimization, can also be applied to other accident reconstructions, and the significant order of initial variables' effects on objectives can provide recommendations for further reconstructions.