基于人工神经网络的驾驶员操纵行为模型

Research on pilot behavior model based on artificial neural network

驾驶员操纵行为受到自身、外界环境和被控对象等多方面的影响,因此驾驶员模型具有非线性的特征。神经网络模型克服了拟线性模型不能反映驾驶员非线性操纵的问题。为了获得建立模型的数据,利用地面模拟器使飞行员对一系列指令进行精确跟踪。获得的指令、飞机状态和驾驶员输入信息等参数即可作为神经网络模型的训练样本。神经网络驾驶员模型的训练和测试结果表明,该建模方法是合理、准确的,可以应用于人机闭环系统中驾驶员操纵的研究。

Pilot control behavior will be affected by the motion of object,pilot-himself and environments,all these disturbances make the control behavior present non-linear characteristics.Therefore the pilot model should have these features which the quasi-linear model could not represent,but the artificial neural network（ANN） pilot model has this pilot control behavior features.The test on ground-based simulator made the flight and pilot＇s data to use for making the ANN pilot model.Compared with McRuer pilot model,the precision and applicability of neural network pilot model could be better.