Robust Approaches to Long-Term Maneuvering-Target Tracking
Abstract
Two complimentary approaches to long-term tracking of a maneuvering target, which are robust to changes in both target maneuver and measurement noise characteristics, are derived in this report. The first approach uses a hidden Gauss-Markov model (HGMM) to estimate the target maneuver and measurement noise characteristics at each update, and produces the track estimate as an ancillary output. The second approach uses continuous Gaussian mixture models (CGMMs) to accommodate heavy-tailed, non-Gaussian behavior in these characteristics at each update, and produces the track estimate as the primary output. Both approaches reduce to iterative Kalman smoothing problems in the linear case.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 11, 2019
- Accession Number
- AD1103353
Entities
People
- Marcus L. Graham
- Michael J. Walsh
Organizations
- Naval Undersea Warfare Center