Distributed Joint Probabilistic Data Association Filter With Hybrid Fusion Strategy

Abstract

This paper investigates the problem of distributed multi-target tracking over a large-scale sensor network, consisting of low-cost sensors. Each local sensor runs a joint probabilistic data association filter to obtain local estimates and communicates with its neighbours for information fusion. The conventional fusion strategies, i.e., consensus on measurement and consensus on information, are extended to multi-target tracking scenarios. This means that data association uncertainty and sensor fusion problems are solved simultaneously. Motivated by the complementary characteristics of these two different fusion approaches, a novel distributed multi-target tracking algorithm using a hybrid fusion strategy, e.g., a mix between consensus on measurement and consensus on information, is proposed. A distributed counting algorithm is incorporated into the tracker to provide the knowledge of the total number of sensor nodes. The new algorithm developed shows advantages in preserving boundedness of local estimates, guaranteeing global convergence to the optimal centralised version and being implemented without requiring no global information, compared with other fusion approaches. Simulations clearly demonstrate the characteristics and tracking performance of the proposed algorithm. Index Terms Multi-target tracking, Multi-sensor fusion, Distributed fusion, Joint probabilistic data association, Hybrid fusion I. INTRODUCTION Wireless sensor networks have attracted great attention in recent decades thanks to their critical importance in a wide range of applications, including environmental monitoring [1], ground vehicle tracking [2]-[4], air traffic control [5], spacecraft navigation [6], vision-based pedestrian tracking [7], etc. The availability of low-cost sensors has enabled employment of multiple sensor nodes to large-scale sensing tasks [8]. Low-cost sensors, however, are generally subject to high clutter rate and low detection probability, leading to performance degradation, especially in multi-target tracking (MTT) scenarios [9]. Leveraging proper fusion algorithms over the sensor network could counteract the drawbacks of low-cost sensors and thus enhance the tracking performance. To this end, this paper aims to address the problem of distributed MTT in a sensor network. The multi-sensor data integration or fusion can be categorised into three architectures in general: centralised, decentralised and distributed [10], [11], as shown in Fig. 1. The centralised fusion architecture simultaneously Shaoming He, Hyo-Sang Shin and Antonios Tsourdos are with the School of Aerospace, Transport and Manufacturing, IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2 processes the measurements provided by all sensors in a fusion centre, which directly connects with all sensor nodes. Although data fusion through a fusion centre is ideally Bayesian optimal, the fusion centre cannot effectively communicate with all sensors for large-scale sensor networks because of physical constraints, e.g., communication delay, limiting communic...

Review of Data Fusion Methods for Real-Time and Multi-Sensor Traffic Flow Analysis

“…One of KF implementation benefits is that it does not require much data to be kept in the memory, and the only critical data is the previous state of the sensor signal. This issue proves that the KF suits real-time application systems with a minimum system specification requirement [91]. However, KF is limited to linear models of domain problems [92].…”

Distributed hybrid consensus–based square-root cubature quadrature information filter and its application to maneuvering target tracking

“…However, in practical applications, only a finite, even small number of consensus iterations are accessible. In such a case, the choice with l i, k = N may cause some nodes to overestimate the novel information, 3,23,37 which should be avoided to preserve Input: The prior information messageŷ i, kjkÀ1 , S i, y, kjkÀ1 n o at time instant k, the consensus weights p ij , and the total number of consensus iterations L.…”

“…However, in practical applications, only a finite, even small number of consensus iterations are accessible. In such a case, the choice with λ i , k = N may cause some nodes to overestimate the novel information, 3,23,37 which should be avoided to preserve consistency of the estimates. An alternative solution can be exploited to compute a normalization factor in a distributed manner, which is capable of improving the filtering performance as well as maintaining consistency of local estimates.…”

Data Association Methods via Video Signal Processing in Imperfect Tracking Scenarios: A Review and Evaluation

“…Although there are many data association methods [15][16][17][18][19][20], how to build a more efficient data association method is still one of the research difficulties. e range of association of MOT data mentioned in the previous paper [21][22][23][24][25] is very wide, and most of them deal with the information collected by sensors, such as different forms of object track information collected by radar, sonar, etc., after pairing and comparing the object track tracked by sensors with the predicted object track, and the correct object track is finally determined [26][27][28][29]. e paper focuses on the data association method based on video MOT, which only deals with video information and no other sensors.…”