Open AccessNon‐myopic sensor scheduling to track multiple reactive targets
Author(s) -
Zhang Zining,
Shan Ganlin
Publication year - 2015
Publication title -
iet signal processing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.384
H-Index - 42
eISSN - 1751-9683
pISSN - 1751-9675
DOI - 10.1049/iet-spr.2013.0187
Subject(s) - computer science , kalman filter , scheduling (production processes) , markov decision process , partially observable markov decision process , real time computing , interception , markov process , mathematical optimization , markov chain , artificial intelligence , markov model , machine learning , mathematics , ecology , statistics , biology
This study addresses the sensor scheduling problem of selecting and assigning sensors dynamically for multi‐target tracking. The authors goal is to trade off the tracking accuracy and the interception risk in a period of time. The interception risk is incurred by the fact that the emission energy originating from a sensor can be intercepted by the target during the tracking mission. To react to sensor emission, the targets are able to switch between dynamic models. This non‐myopic sensor scheduling problem is formulated as a partially observable Markov decision process, where the one‐step reward is constructed by combining the tracking error with the interception probability and the information state is tracked by the interacting multiple model extended Kalman filtering. A novel sampling approach using the unscented transformation is proposed for long‐term reward approximation. Numerical simulations illustrate the validity of the proposed scheduling scheme.