基于粒子群优化人工神经网络的临界行车安全距离预测

Prediction and Simulation of Critical Driving Safety Distance Based on PSO-ANN

针对行车安全距离预测中的各种非线性问题,提出一种基于粒子群优化人工神经网络(PSO-ANN)的临界安全距离预测方法。通过粒子群优化(PSO)算法优化人工神经网络(ANN)的权值和阈值,避免ANN容易陷入局部最优的问题,并通过迭代找到全局最优解。以路面情况、前后车速度以及前车减速度作为输入,临界行车安全距离作为输出,应用PSOANN建立预测模型,通过训练收集的样本数据预测行车安全距离,并与当前常用的ANN预测结果进行比较,结果表明:与ANN方法相比,PSO-ANN算法更稳定,且预测结果的平均绝对误差降低了7.8%。

To solve various nonlinear problems in predicting safety distance, a prediction method for is proposed critical driving safety distance based on Particle Wwarm Optimization and Artificial Neural Network(PSO-ANN). PSO is used to optimize weight and threshold of ANN, which prevents ANN from easily falling into local optimum, and finds the global optimal solution through iteration. With road condition, the speed of the preceding vehicle, the speed of following vehicle, and the deceleration of preceding vehicle as input, critical driving safety distance as output, a prediction model with PSO-ANN is built. By training the collected sample data, the prediction result of driving safety distance is obtained,and compared with the prediction result of ANN is built. The experimental results show that the PSO-ANN algorithm is more stable and the average absolute error of prediction results is reduced by 7.8% compared with ANN method.