基于禁忌搜索算法的换道轨迹优化

The Optimization of Lane Changing Trajectory Based on Tabu Search Algorithm

针对换挡过程的易变性,提出利用禁忌搜索算法对换道轨迹进行优化。在真车试验数据分析的基础上,对换道过程进行阶段划分。根据换道过程中驾驶员可能的操作将驾驶员操作行为,划分为12种操作模式;并依据不同时窗长度下换道过程划分准确率的高低,确定出最佳的时窗长度;并建立相应的换道轨迹预测方程。根据换道轨迹预测方程并结合车辆当前的运动状态对下一时刻的状态做出估计;并实时估算换道所需时间。将换道完成所需时间作为适应度函数,以下一时段可能出现的状态序列为候选解,通过设定禁忌表、禁忌长度以及初始解向量对换道轨迹进行实时优化。通过测试不难发现,禁忌搜索算法的收敛速度更快,确定的换道轨迹是全局最优。

Because lane changing process exists variability, tabu search algorithm is applied to optimize lane changing trajectory. The stage division of lane changing process is achieved on the basis of the analysis of the real car test datas. The driver behaviors are divided into 12 kinds of operation modes according to the possible operations driver adopts during lane changing. The best time window length can be determined according to the classification accuracy of lane changing process under different time window lengths. And then the corresponding lane changing trajectory prediction equation is established. Lane changing trajectory prediction equation combined with the current motion state of vehicles are adopted to estimate the state next moment. The duration of lane changing is achieved real-time estimation. The time that lane changing need is as fitness function. The possible state sequences appear in the following a period of time are taken as the candidate solutions. The lane changing trajectory is optimized in real time by setting tabu table, tabu length and the initial solution vector. It can be seen from the test result that convergence speed of tabu search algorithm is faster and the determined changing trajectory is the global optimal.