一种预测驾驶员操纵行为的建模方法

Approach of pilot modeling for predicting pilot control behavior

在飞机设计中,应用驾驶员数学模型预测飞机飞行品质是避免人机系统出现不良耦合的重要途径之一.驾驶员神经网络模型是利用飞行模拟实验数据建立的驾驶员模型.基于该模型,针对纵向单通道俯仰跟踪任务,详细讨论了具有不同增益、不同短周期振荡频率飞机构型的驾驶员操纵行为变化规律.研究结果表明:对于特性相似的飞机构型,其驾驶员操纵特性也相似.因此,提出了一种利用相似构型的驾驶员操纵行为特性建立驾驶员预测模型的方法.通过对预测模型进行精度评价,可以证明采用本方法能够获得比较满意的驾驶员操纵行为特性预测结果.

During the process of aircraft design,the mathematical model of pilot control behavior characteristics is always used to predict aircraft flying qualities.This is one of the important methods to avoid pilot-aircraft adverse coupling.The pilot neural networks（NN） model was derived from experimental data obtained during pilot-aided tests using a special workstation.For a set of different aircraft dynamics configurations,which had different amplifications and different short period mode natural frequency,the pilot control behavior characteristics in the single-loop longitudinal pitch tracking task were studied based on neural networks model.The result shows that pilots have similar control behavior for similar aircraft dynamics configurations.Therefore,an approach used pilot control behavior of similar aircraft configurations to build a pilot model to predict pilot control behavior of new configuration was suggested.This pilot model＇s precision was further calculated.And the result verifies applicability of this new approach for pilot predicting modeling.