Research on the Evaluation of Lane-changing Models Based on Improved TOPSIS

An improved evaluation method based on the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) is proposed. This method incorporates the comprehensive weights of indicators to assess high-speed lane-changing models of self-driving cars. Firstly, based on safety, comfort, and efficiency, six indicators, including time to collision, collision margin, lateral acceleration, longitudinal acceleration, lane-changing time, and lane-changing distance, are selected to establish an evaluation system for lane-changing models. Secondly, the entropy weight method and hierarchical analysis method are used to calculate the objective weight and subjective weight of the evaluation indexes. The TOPSIS evaluation model is improved based on the calculated integrated weight. Finally, the TOPSIS model is used to evaluate the performance of different lane-changing models in different simulation environments. The results indicate that the lane change evaluation model integrates subjective experience and objective data information, reduces information overlap between indicators, and ensure that the evaluation results are made more reasonable.