基于时空特征与交互关系建模的换道意图预测
高帅帅,魏诚,惠飞,张竟成,宋汉辰

Lane-changing intention prediction based on spatiotemporal feature and interaction modeling
Shuaishuai GAO,Cheng WEI,Fei HUI,Jingcheng ZHANG,Hanchen SONG
表 4 输入长度为2 s时不同模型的性能对比
Tab.4 Performance comparison of different models with 2 s input length
模型类别$ P $/%$ R $/%F1 /%$ {A}_{\text{cc}} $/%
Social-LSTM
(模型A)
LK87.1794.8190.8393.46
LLC96.9894.3995.67
RLC97.1091.1994.05
ST-GCN(模型B)LK91.6996.5994.0895.94
LLC98.7494.5596.60
RLC97.7796.7097.23
Transformer
(模型C)
LK95.3096.7596.0297.33
LLC98.2399.1698.70
RLC98.4996.0797.27
GATv2
(模型D)
LK71.2478.3474.6380.53
LLC86.4085.3285.86
RLC85.3577.9281.47
Transformer-GATv2
(模型E)
LK97.5196.5997.0598.04
LLC98.5499.0698.80
RLC98.0798.4898.27