Multiple model box-particle cardinality balanced multi-target multi-Bernoulli filter for multiple maneuvering targets tracking

Abstract

Cardinality balanced multi-target multi-Bernoulli (CBMeMBer) filter has been proved as a promising method in the context of multi-target tracking with an unknown number of targets, clutter and false alarms. For tracking maneuvering targets, the CBMeMBer filter has been extended by using jump Markov models (JMM). However, the standard particle implementation of the multiple model CBMeMBer (MM-CBMeMBer) filter requires a large number of particles in order to obtain a satisfactory performance. Based on the capability of box-particle filter to process measurements which are affected by bounded errors of unknown distributions and biases, a box-particle implementation of the MM-CBMeMBer filter is proposed. Simulation result shows that the proposed MM-Box-CBMeMBer filter can obtain similar accuracy results with a MM-Particle-CBMeMBer filter but considerably reduce the computational costs. Meanwhile, in the presence of strongly biased measurements, it is shown that the MM-Box-CBMeMBer filter is superior to the MM-Particle-CBMeMBer filter.