跟驰模型场景基准分析

Scenario-based Benchmark Analysis for Car-following Models

摘要:

设计包含7种驾驶工况的自由流场景及跟驰场景，对MITSIM（microscopic traffic simulator）、Gipps模型、Wiedemann模型、FVD（full velocity difference）模型、IDM（intelligent driver model）、S?K模型及LCM（longitudinal control model）7个不同机理下典型跟驰模型进行基准测试。结果表明：各模型均可在一定程度上完成自由流场景下各工况测试，其中LCM的表现最符合认知，IDM的加速度最小且启动时间最长；仅Gipps模型、FVD模型、IDM及S?K模型可完成跟驰场景下各工况测试，其中FVD模型对于前车的状态变化具有最及时的反应；S?K模型在跟驰场景各工况下的速度与加速度始终处于微小振荡中，这也较为符合实际驾驶情景。

Abstract:

Seven typical car-following models under different mechanisms， including microscopic traffic simulator（MITSIM）、Gipps model、Wiedemann model、full velocity difference（FVD） model、intelligent driver model（IDM）、S-K model and longitudinal control model（LCM）， were tested under seven designed driving regimes. It is shown that each model is able to complete three free flow regimes to a certain extent， and the performance of LCM is consistent with the cognition rather than the other models. IDM has the smallest value of acceleration and the longest acceleration process. Only Gipps model， FVD model， IDM， and S-K model are able to complete four car-following regimes， among which FVD model has the most timely response to the state change of preceding vehicle. The speed and acceleration of S-K model are always in a slight oscillation under seven driving regimes， which is in line with the actual driving situation.