GNSS/INS Semi-deep Integration with Federated Filtering for High Dynamic Vehicle

摘要

Deep integration of GNSS/INS is an advisable solution to improvehigh dynamic vehicle’s navigation accuracy and reliability. The key featureof deep integration is that the navigation solution is fed back into receiverbaseband to improve GNSS signal tracking performance. However, deeplyintegrating all-in-view satellites from multiple GNSS constellations withINS will lead to heavy computational burden and fault cross-infectionamong receiver channels. To address the issue, this paper investigates afederated filtering algorithm based semi-deep integration of GNSS/INS.The available GNSS signals are divided into two groups on the basis of thelevel of dynamics and C/N0 (Carrier-to-Noise Ratio). One group of theGNSS signals are deeply coupled with INS as traditional deep integration,the remaining signals are tracked without INS feedback but tightly coupledwith INS. Estimates of these two group integrations are eventually fused bya federated filter with designed information-sharing coefficients. Thus,these two groups of GNSS signal tracking as well as INS can benefit frommutual assistance while the system enjoying less complexity and high faulttolerance. The scheme effectiveness is verified by a simulation case underhigh dynamic scenario. Results show that compared with traditional deepintegration, the presented integration has competitive efficiency with littleaccuracy degradation and leads to an easier implementation and parametertuning.