基于支持向量机的概率型飞行冲突探测算法

A probabilistic flight conflict detection algorithm based on SVM

为解决自由飞行条件下冲突探测概率估计问题,提出了一种基于支持向量机(SVM)的概率型飞行冲突探测算法。首先对飞机运动轨迹及保护区建模,飞机预测航迹误差服从高斯分布,以目标机为中心建立椭球保护区E模型,获得冲突和不冲突样本。在此基础上,选取飞机当前位置、速度矢量和向前看时间为特征量,采用SMOTE重采样方法解决样本不均衡问题,进行SVM训练,并通过Sigmoid函数概率映射法求解冲突概率,建立冲突探测模型。仿真结果表明,所提出的算法能够探测中短期向前看时间内飞行冲突,所得冲突概率准确、虚警率较低。

For the purpose to predict flight conflict in free flight,a probabilistic conflict detection algorithm based on support vector machine(SVM)was proposed.Firstly,trajectory prediction and protected zone model were constructed.An ellipsoidal protection zone which centered on target aircraft was established,and trajectory prediction errors were modeled as Gaussian distribution,so that training samples were collected.Secondly,the flight conflict detection model was trained in SVM:The current positions,velocity vectors and predicted time were extracted as characteristic quantities,meanwhile,SMOTE method was introduced to handle with imbalanced data sets.Afterwards,the samples were trained in SVM.Finally,the conflict within look-ahead time in medium and short terms would be predicted.Moreover,the conflict probability was solved by the Sigmoid function mapping method.In the simulations,it is proved that the flight conflict can be detected by this method,and the conflict probability results are accurate,and the false alarm rate is low.