基于神经网络的模型参考自适应逆飞行控制

针对飞行安全控制问题,结合动态逆方法和神经网络理论,提出了一种基于改进BP神经网络的模型参考自适应逆控制方法,应用到飞行控制系统中。该方法在控制器中引入神经网络算法,在经典BP神经网络控制算法的基础上,使用遗传蚁群算法优化神经网络参数,在线调整网络的权值和阈值,避免了传统梯度下降法的缺点,提高了自适应算法的效率,达到了抗干扰的目的。从而改善了飞机飞行稳定性和操纵性。在波音747—100/200飞机模型上仿真实验表明了该方法的可行性和鲁棒性,能够保障飞行安全。

Neural Networks Based Model Reference Adaptive Dynamic Inversion Flight Control

According to the flight safety control problem, there presents a design for model reference adaptive control based on improved BP neural networks together with dynamic inversion method and neural networks. Traditional BP neural network based, introduce the neural networks algorithm, applying Genetic Algorithm-Ant Colony Algorithm(GA-ACA)to optimize the parameter of neural networks, and adjusting weights and thresholds online,avoid the traditional Gradient Descent shortcomings. It increases the efficiency of the adaptive algorithm and achieves the anti-interference purpose, improves the flight stability and dirigibility.The simulation experiment of Boeing 747-100/200 flying model proved its robustness ability and it can guarantee the safety of flight.