A low-complexity interacting multiple model filter for maneuvering target tracking

Syed Safwan Khalid, Shafayat Abrar

Research output: Contribution to journalArticlepeer-review

15 Scopus citations

Abstract

In this work, we address the target tracking problem for a coordinate-decoupled Markovian jump-mean-acceleration based maneuvering mobility model. A novel low-complexity alternative to the conventional interacting multiple model (IMM) filter is proposed for this class of mobility models. The proposed tracking algorithm utilizes a bank of interacting filters where the interactions are limited to the mixing of the mean estimates, and it exploits a fixed off-line computed Kalman gain matrix for the entire filter bank. Consequently, the proposed filter does not require matrix inversions during on-line operation which significantly reduces its complexity. Simulation results show that the performance of the low-complexity proposed scheme remains comparable to that of the traditional (highly-complex) IMM filter. Furthermore, we derive analytical expressions that iteratively evaluate the transient and steady-state performance of the proposed scheme, and establish the conditions that ensure the stability of the proposed filter. The analytical findings are in close accordance with the simulated results.
Original languageEnglish (US)
Pages (from-to)157-164
Number of pages8
JournalAEU - International Journal of Electronics and Communications
Volume73
DOIs
StatePublished - Jan 22 2017

Fingerprint

Dive into the research topics of 'A low-complexity interacting multiple model filter for maneuvering target tracking'. Together they form a unique fingerprint.

Cite this