Abstract
A method named FDER is proposed to discriminate unreliable results, reduce their probability and finally fuse multi-model results. The FDER first determines sub-intervals of the final output result, then the evaluation matrix and the model reliability are calculated, the minimum model reliability means that the model and its corresponding result may be unreliable. Finally, the probability of each sub-interval is calculated. All sub-intervals and their corresponding probability are considered as the final fused result. Results of the numerical analysis demonstrate that the fused result of the FDER is very close to the existing method when all accident reconstruction results are reliable, and the final fused results still follow a normal distribution when there are problematic results in those reconstruction results. The applicability of the proposed method is demonstrated by analyzing an actual vehicle-motorcycle accident. The result show that the FDER works well in practice.
Disclosure statement
No potential conflict of interest was reported by the authors.