交通车辆换道行为预测方法

Research on a prediction method of lane changing operations for vehicles

为了准确预测高级驾驶辅助系统目标车辆换道行为,提出一种换道概率累积分布函数的车辆换道行为预测方法。对传感器采集的大量车辆换道与道路保持数据进行统计分析,针对换道行为的2种实时特征信号和4种阈值信号,拟合出计算换道概率的累积分布函数。针对实时特征信号与阈值特征信号所表达的不同概率,构建了正常换道与复杂换道等16种驾驶场景,根据实际驾驶环境进行了概率权重系数分配,最终确定目标车左、右换道行为概率。基于交通车辆实际驾驶数据,针对目标车正常换道场景、压线换道场景以及不连续换道等复杂场景,验证了车辆换道行为概率累积分布函数预测方法的有效性。结果表明,该方法能够准确预测车辆实际驾驶环境下的换道意图,实时性与鲁棒性好,适合于实车应用。

In order to accurately predict lane changing operations of the target vehicle in Advanced Driving Assistance System,this paper proposes a prediction method of vehicle lane changing based on a cumulative distribution function of lane changing probability.Statistical analysis of a large number of vehicle route changing and road maintenance data collected by sensors is performed.For two real-time signals and four threshold signals of route changing,the cumulative distribution function for calculating lane changing probability is fitted.Sixteen driving scenarios,such as normal and complex lane changing,are constructed according to the different probabilities expressed by real-time and threshold signals.The probability weight coefficients are allocated according to the actual driving environment,and the left or right lane changing probabilities of the target vehicle are finally determined.Based on the actual driving data of vehicles,the validity of the cumulative distribution function of lane changing probability is verified in complex scenarios,such as normal lane changing,cross-line lane changing and discontinuous lane changing of the target vehicle.The results show that the method can accurately predict the driver’s lane changing intentions in an actual driving environment,and has good real-time and robustness,which is suitable for real cars.