神经网络与贝叶斯滤波器在换道预测中的应用

The Application of Bayesian Filter and Neural Networks in Lane Changing Prediction

为提高车辆在换道过程中的行车安全性。提出一种基于BP神经网络与贝叶斯滤波器的换道意图预测方法,通过车道线传感器、方向盘转角传感器和车身CAN总线采集相关表征参数,将其作为BP神经网络输入数据,对驾驶人换道意图进行初步预测,BP神经网络输出结果作为贝叶斯滤波器输入数据,对BP神经网络预测结果作进一步修正。对模型利用真实换道数据进行训练和检测,结果表明此模型的预测准确率达到91.38%,相较于单一的BP神经网络模型,预测准确率提高了6%,并且具有更强的通用性。

In order to improve safety during lane change, we proposed lane change intent prediction method based on neural networks and Bayesian filters. The method uses the lane line sensor, steering wheel angle sensor and in-vehicle CAN bus acquisition characterization parameters. The above acquisition parameters as the neural network input data, driver’ s lane change intention preliminary forecast, take the output of BP neural network as the input of Bayesian filters, and then amendments the results of BP neural network. Using real vehicle lane changing data training and testing the model. The results show that prediction accuracy rate of BP neural network and Bayes-ian filters reaches 91. 38%. The forecast accuracy increased by 6 percentage points compare to single BP neural network and has better versatility.