平流层飞艇升降阶段BPNN惯性导航算法

软式平流层飞艇艇体在上升和下降时经常呈堆叠状态，GPS信号会被艇体间歇性遮挡，因而只能采用惯性导航。为保证在飞艇上升和下降过程中，INS/GPS组合导航系统在被艇体遮挡GPS时仍能够提供满足精度要求的导航信息，设计了一种改进的反向传播神经网络(Back Propagation Neural Network, BPNN）惯性导航算法。采用神经网络，根据惯性导航系统在1s内的速度均值和姿态变化量，预估其在1s末的位置误差和速度误差，并对惯性导航结果进行修正。仿真实验和跑车试验结果表明，在GPS失效的30s内，新算法使得位置误差低于15m，速度误差低于0.7m/s，误差相比纯惯性导航降低了85%。

BPNN Inertial Navigation Algorithm for Stratospheric Airship Climb and Descent Phase

The skeleton-free stratospheric airship hull is often stacked by the wind during the climbing and descending phase and GPS signal is often intermittently blocked by the hull, which leads that only inertial navigation can be used and the navigation errors rise up quickly. In order to ensure that the INS/GPS integrated navigation system can still provide navigation information that meets the accuracy requirements in this situation, an improved BPNN inertial navigation algorithm is designed, which uses the neural network to estimate the position error and velocity error at the end of 1s according to the mean velocity and attitude change within 1s of inertial navigation, and corrects the inertial navigation results. Simulation and road testing results show that the position error is less than 15m and the velocity error is less than 0.7m/s within the 30s of GPS failure, which is 85% lower than the inertial guidance.