An event-triggered distributed Mδ-GLMB filter

The marginalized $\delta$-generalized labeled multi-Bernoulli (M$\delta$-GLMB) filter has demonstrated its effectiveness in multi-target tracking, and fusion rules have been proposed for M$\delta$-GLMB densities so as to allow its use in distributed multi-target tracking applications. However, the M$\delta$-GLMB density is formed based on hypotheses whose numbers increase exponentially with respect to the cardinality of the label set, thus imposing a heavy communication burden on the sensor network. To overcome this problem, two event-triggered (ET) strategies are devised in this paper for fusion of M$\delta$-GLMB densities, which are able to significantly reduce the data exchange rate at the price of a slight performance loss. Specifically, a method for tuning the hypothesis weights is proposed for the ET strategy so as to guarantee the normalization of the M$\delta$-GLMB density. The effectiveness of the proposed methods is verified via simulation results.