基于GA-BPNN的航空器无人驾驶车引导跟驰模型  

GA-BPNN Based Unmanned Guided Aerial Vehicle Gliding Heeling Behavior Research

作　　者：赵庆 袁德周 张天雄 朱新平[1] ZHAO Qing;YUAN De-zhou;ZHANG Tian-xiong;ZHU Xin-ping(Civil Aviation Flight University of China,GuangHan 618000,China)

机构地区：[1]中国民用航空飞行学院,四川广汉618000

出　　处：《航空计算技术》2024年第6期43-49,共7页Aeronautical Computing Technique

基　　金：中央高校基本科研业务费专项资金项目资助(J2023-047);国家重点研发计划项目资助(2022YFB2602004)。

摘　　要：本研究将无人驾驶技术引入机场飞行区的场面运行中,特别关注无人驾驶车(UGV)在主滑行道上引导航空器的跟驰模型。针对UGV与传统有人驾驶航空器在运动学特性上的差异,分别为其建立了梯形和S型运动学模型。为提高速度预测的准确性,构建了一个基于神经网络的速度优化模型。鉴于传统神经网络在隐藏层回归系数上存在较大偏差,引入了基于遗传算法优化的BP神经网络(GA-BPNN)。在鄂州花湖机场进行的仿真实验结果表明,GA-BPNN在速度误差、航空器预测速度波动范围以及无人驾驶车辆预测速度稳定性方面均优于传统方法。GA-BPNN的速度误差小于1%,航空器预测速度波动范围控制在±2 m/s以内,无人驾驶车辆预测速度的动态误差小于0.5%。This study introduces the integration of unmanned driving technology into airport flight zone operations,focusing on the platooning model between Unmanned Ground Vehicles(UGVs)and manned aircraft on main taxiways.Acknowledging the kinematic differences between UGVs and manned aircraft,we developed distinct trapezoidal and S-type kinematic models to characterize their motion.To enhance velocity prediction accuracy,a neural network-based velocity optimization model was established.Given the substantial deviation in the regression coefficients of the hidden layer in traditional Neural Networks(NN),we introduced a Genetic Algorithm-optimized Backpropagation Neural Network(GA-BPNN).Simulation experiments conducted at Ezhou Huahu Airport validated the superiority of GA-BPNN in velocity error,aircraft predicted velocity fluctuation range,and stability of unmanned vehicle velocity prediction.The GA-BPNN achieved a velocity error of less than 1%,with aircraft predicted velocity fluctuations within±2 m/s,and unmanned vehicle velocity prediction stability with a motion error of less than 0.5%.

关 键 词：跟驰行为 无人驾驶车 遗传算法 神经网络 速度预测 

分 类 号：O224[理学—运筹学与控制论]