融合动态风险图与多变量注意力机制的车辆轨迹预测模型
陈文强,冯琳越,王东丹,顾玉磊,赵轩

Vehicle trajectory prediction model integrating dynamic risk map and multivariate attention mechanism
Wenqiang CHEN,Linyue FENG,Dongdan WANG,Yulei GU,Xuan ZHAO
表 1 不同模型的均方根误差对比
Tab.1 Root mean square error comparison of different models
数据集tp/sRMSE/m
CS-LSTMMHA-LSTMSTDANiNATranSTGAMTGRIPGSTCNRGMA
NGSIM10.610.410.420.410.210.370.440.16
NGSIM21.271.011.0110.780.860.830.60
NGSIM32.091.741.691.71.491.451.331.17
NGSIM43.12.672.562.572.42.212.011.88
NGSIM54.373.833.673.663.573.162.982.73
NGSIM平均值2.291.931.871.871.691.611.521.31
HighD10.220.060.190.040.070.05
HighD20.610.090.270.050.190.14
HighD31.240.240.480.210.320.27
HighD42.10.590.910.540.610.47
HighD53.271.181.661.111.140.80
HighD平均值1.490.430.700.390.470.35