驾驶员神经网络模型与频域拟线性模型的比较研究

 针对在双通道人—机控制系统中,由于驾驶员在各通道间的注意力分配,造成操纵行为具有明显的非线性这一特性,提出了用神经网络模型进行驾驶员行为描述的新方法。通过在飞行模拟器上进行双通道控制的人—机闭环仿真实验,分别对驾驶员的神经网络模型和频域拟线性模型进行识别,并对两种模型的识别结果进行了精度分析和仿真验证。结果表明神经网络模型在更宽的频带范围内具有良好的精度,而频域拟线性模型只在低频范围内精度好,其频带范围相对较窄。这说明神经网络模型在描述驾驶员非线性行为特征方面优于传统的频域拟线性模型。

Comparison of Neural Network Model and Freguency Domain Quasi-Linear Model for Human Pilots

In a previous study, it was found that the pilot behavior shows obviously nonlinear characters in the multi-loop pilot-vehicle system. Therefore, a new method, neural-networks, is focused on the pilot modeling. A real time simulation is carried out on a simulator, in which the pilot traces the command velocity and pitch angle of aircraft by controlling the stick and throttle respectively. The neural network model is identified and its precision expression is further proposed. The result obtained verifies the validity of the proposed models compared with the quasi-linear model. It is proved that the neural network model is an effective model to describe the nonlinear behavior of pilots.