Event-Triggered Distributed Multitarget Tracking

Abstract

This paper focuses on reducing communication bandwidth and, consequently, energy consumption in the context of distributed multitarget tracking (DMT) over a peer-to-peer sensor network. A consensus cardinalized probability hypothesis density (CCPHD) filter with event-triggered communication is developed by enforcing each node to broadcast its local informa- tion to the neighbors only when it is worth to, i.e. the node has gained a sufficient amount of information with respect to its latest broadcasting. To this end, each sensor node separately evaluates the discrepancies of the cardinality probability mass function (PMF) and of the spatial probability density function (PDF) between the current local posterior and the one recoverable from neighbors after the latest transmission. Then, each sensor node selectively sends the specific information on the multitarget distribution (i.e. the cardinality PMF or the spatial PDF or both) that is considered to be worth transmitting (i.e., such that the respective discrepancy exceeds a preset threshold). Two types of discrepancy measures, i.e. Kullback-Leibler divergence (KLD) and Cauchy-Schwarz divergence (CSD), are investigated. The performance of the proposed event-triggered CCPHD filter is evaluated through simulation experiments.