Getting The Best of Particle and Kalman Filters: GNSS Sensor Fusion using Rao-Blackwellized Particle Filter
Shubh Gupta, Adyasha D. Mohanty, Grace Gao · Proceedings of the Satellite Division's International Technical Meeting (Online)/Proceedings of the Satellite Division's International Technical Meeting (CD-ROM) · 2022
In this paper, we develop a hybrid Bayesian filter for GNSS and visual odometry (VO) fusion in urban environments that combines the tracking efficiency of Kalman filter with the superior uncertainty modeling of particle filter. Our filter design employs Rao-Blackwellization to decouple the state into a non-linearly tracked position and linearly tracked orientation, velocity and carrier phase integer ambiguities. This factorization allows our filter to efficiently track the state along with a rich probability distribution of the position. Moreover, we utilize the tracked position probability distribution to quantify the uncertainty in situations where the tracking is erroneous. We evaluate our approach for state and uncertainty estimation using GNSS-VO fusion on real-world data and demonstrate that our filter exhibits comparable computation efficiency and improved state and uncertainty estimates over other Bayesian filter baselines.