Consideration of Particle Flow Filter Implementations and Biases
Abstract
Particle flow filters are appealing due to their potential resistance to particle collapse. However, common implementations exhibit undesirable biases or particle divergence. This paper shows that the explicit, incompressible, and diagonal flows, unlike the Gromov flow, are inherently biased. Another issue is errors in the numerical integration of the flow. The benefits of implicit stochastic-integration methods are demonstrated and a new adaptive step-size selection heuristic is presented.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 11, 2020
- Accession Number
- AD1091426
Entities
People
- Codie Lewis
- David F. Crouse
Organizations
- United States Naval Research Laboratory