面向多元未知环境的基于深度高斯过程组合导航轨迹预测方法

Integrated navigation trajectory prediction method based on deep Gaussian process for multiple unknown environments

传统惯导/卫导组合导航在多元复杂环境下易受干扰,从而导致观测量异常影响导航性能。以无人驾驶车辆为研究对象,展开提升组合导航系统导航精度的研究。采用深度高斯过程(deep Gaussian process,DGP)辅助估计位置的方法减小组合导航误差,提高定位性能。基于DGP的辅助导航方法不仅可以预测无人驾驶车辆的标称轨迹,同时可以预测各时刻位置可信区间的概率分布,为基于深度学习模型的数据融合预测方法提供了严格的理论解释性。真实历史数据下的多重对比实验表明,该算法较传统深度神经网络算法具有更高的精度和可靠性。基于DGP的辅助导航方式能有效提高全球卫星定位系统信号失锁时的导航模型性能,实验表明相对于纯惯性导航系统(integral navigation system,INS)解算和长短期记忆(long and short term memory,LSTM)进行导航信号补偿定位精度分别提高了97.32%和52.13%。

The traditional inertial navigation and satellite navigation integrated is prone to interfered in complex multi-element and hybrid environment,which affects navigation performance and leads to abnormal observations.We take unmanned ground vehicle as the object and carry out the research on improving the accuracy of the integrated navigation system.A deep Gaussian process(DGP)assisting location estimation method is used to reduce the integrated navigation error and improve the positioning performance.The assisted navigation method based on DGP can not only predict the trajectory of the unmanned vehicle,but also can estimate the probability distribution of the position confidence interval at every time,which provides strict theoretical interpretation for the data fusion prediction method based on deep learning model.Multiple comparison experiments with real historical data show that the proposed framework achieves higher accuracy and reliability than deep neural network algorithms.The DGP-based auxiliary navigation can effectively improve the performance of the navigation model when the global positioning system(GPS)signal is out of lock,and the experiments show that the navigation signal compensation positioning accuracy is improved by 97.32%and 52.13%respectively compared with pure integral navigation system(INS)and long and short term memory(LSTM).