Core Concepts
The author proposes the Multiple Update Particle Filter as an efficient method to update particles in a particle filter when dealing with sharp-peaked likelihood functions from multiple observations. By leveraging prior knowledge of distribution spreads, this method enhances position estimation accuracy.
Abstract
The content introduces the Multiple Update Particle Filter (MU-PF) for efficient position estimation using GNSS pseudorange and carrier phase observations. The MU-PF addresses challenges faced by traditional PFs when likelihood functions exhibit sharp peaks, ensuring accurate state estimation. Experimental results demonstrate superior convergence and accuracy compared to conventional methods.
The study highlights the importance of considering the spread of likelihood functions from multiple observations in particle filters. It introduces a systematic approach to updating particles based on distribution breadth, improving convergence speed and estimation accuracy. The proposed method outperforms existing techniques in urban environments, showcasing its effectiveness in real-world applications.
Evaluation through static and kinematic tests validates the efficacy of the proposed method over normal PF and annealed PF approaches. The MU-PF demonstrates rapid convergence with fewer particles, achieving centimeter-level accuracy even after a single observation step. In kinematic tests using real-world data, the proposed method outperforms conventional RTK-GNSS methods, emphasizing its practical utility for accurate position estimation.
Stats
"The number of particles was set to N = 2000."
"In static experiments utilizing GNSS pseudorange and carrier phase observations..."
"Varying the number of particles at 100, 500, and 1000."
Quotes
"The proposed method achieves convergence to the correct position in nearly one observation step."
"The proposed method exhibits superior convergence performance with fewer particles than the annealed PF."
"The proposed method demonstrates accurate position estimation when compared to conventional RTK-GNSS approaches."