基于空间关联的GNSS自主完好性监测方法研究

Receiver Autonomous Integrity Monitoring (RAIM) is an important means to ensure the integrity of civil aviation navigation. The existing method uses the confidence upper bound of ranging error to achieve the integrity monitoring. The error estimation is too conservative and cannot meet the availability requirements of civil aviation precision approach. Therefore, this project proposes a monitoring method that integrates the spatial correlation of navigation error with the integrity risk to improve the availability of the algorithm. Aiming at the problem of insufficient availability for existing fault detection, a dynamic autonomous detection method based on space-related risk model is proposed. Aiming at the problem of overlapping and indistinguishable feature for multi-fault recognition, a multi-fault hierarchical identification method based on spatial correlation is proposed. To solve the small fault that is hard to detect and identify, a Bayesian positioning method with minimal integrity risk is proposed to improve the robustness of position solution. The performance of our method is evaluated by numerical simulation and real data. The project aims to provide theoretical and methodological support for autonomous integrity monitoring of GNSS multi-constellation navigation in the application of precision approach.

接收机自主完好性监测（RAIM）是保障民用航空导航完好性的重要手段。现有方法采用测距误差的置信上界实现完好性监测，其误差估计过于保守，不能满足民航精密进近引导的可用性要求。因此，本项目提出导航误差空间关联特性与完好性风险融合的监测方法，从而提高算法的可用性。针对现有故障检测可用性不足的问题，提出基于空间关联风险模型的动态自主检测方法；针对多故障识别中特征空间相互交叠、无法区分的难题，提出基于空间关联特性的多故障分层识别方法；针对难以检测和识别的小故障，提出基于最小化完好性风险的贝叶斯定位解算方法，提高位置解算的鲁棒性；结合数值仿真和实测数据验证，对本项目方法的性能进行评估，为面向精密进近引导的GNSS多星座导航自主完好性监测系统提供理论和方法支撑。

本课题面向生命安全领域导航定位完好性监测的迫切需求，开展了完好性风险建模、故障检测和风险包络三个层次的研究。（1）针对现有故障模式在对抗场景下的不足，提出GNSS导航可能面临对抗攻击的完好性风险，通过对抗网络生成欺骗性卫星观测，在欺骗误差大小和欺骗卫星数量受到限制的情况下，依然能够躲避现有完好性监测算法从而误导接收机产生错误的定位解算结果。(2) 针对现有故障检测可用性不足的问题，提出了基于深度神经网络的自主完好性监测方法RAIM-NET，利用全部卫星观测的定位解与部分卫星观测定位解的差值作为特征向量，使用多层长短期存储器（LSTM）建立完好性监测的模型结构，并设计了基于解分离法的正则化项，实验表明，RAIM-NET算法能够提高对小故障的检测率，降低完好性监测的保护级，提高完好性监测算法的可用性。(3) 针对多故障识别中特征空间相互交叠、无法区分的难题，提出了一种基于GNSS双差伪距和UWB融合的完好性监测方法，其故障识别率在不同实验场景下，相比于传统的GRSM方法能够提高6%~42%。然后，针对难以检测和识别的小故障，提出基于最小化完好性风险的位置解算方法，提高位置解算的鲁棒性。