有级转向履带车辆的驾驶员操控行为模型

为解决有级转向履带车辆驾驶员操控经验的表征问题，以采集得到的大量真实驾驶数据为依托，提出一种基于操控基元序列的驾驶员操控行为表征模型。操控基元以高斯混合模型表征，并以有向图的形式完成驾驶员操控基元切换序列的提取与类别辨识；依据不同类别的操控基元序列辨识结果完成了对驾驶数据的重分组，利用隐马尔可夫-高斯混合模型完成每一个类别下的驾驶员操控行为模型训练。结果表明：所提取出的操控基元序列既能完成对驾驶员转向操控基元切换行为的表征，又能实现对轨迹基元类别的合理划分；在给定期望轨迹所对应的预测时域内，驾驶员操控行为模型能够实现特定条件下的驾驶员转向操控量预测，预测平均偏差在4.2%以内。

Driver Control Behavior Model for Tracked Vehicle with Stepped Steering Mechanism

For the representation of driving skills at the control level of tracked vehicle with stepped steering mechanism, a driver control behavior model based on control primitive sequence is proposed by utilizing a large amount of collected real driving data. The control primitives are characterized by Gaussian mixture models, and the digraph is applied to complete the extraction and category identification of the driver control primitives sequence. According to the recognition results of different types of control primitive sequence, the driving data is regrouped, and the hidden Markov model-Gaussian mixture model is used to train the driver control behavior model for each type of sequence. The results show that the extracted control primitive sequence can not only represent the driver's steering control primitives switching behavior, but also realize the reasonable division of the trajectory primitive types. Within the predicted time domain corresponding to a given desired trajectory, the proposed model can predict driver control behavior with an average deviation of 4.2% under certain conditions.