Consistent methods for Decentralised Data Fusion using Particle Filters

Abstract

Abstract-This paper presents two solutions for performing decentralised particle filtering in view of non-linear, nonGaussian tracking in sensor networks. The issue is that no known methods exist to deal with correlated estimation errors due to common past information between two discrete particle sets.The first method transforms the particles to a Gaussian Mixture Model, the second approximates the set by a Parzen density estimate. Both of these representations accommodate consistent fusion and maintain accurate summaries of the particles. Requiring less bandwidth than particle representations, transformations to GMMs or Parzen representations for communication provide an added advantage.The accuracy in which the algorithms summarise the particle set, fusion methods and bandwidth requirements of each representation will be compared. Our results show that whilst less GMM components are required to summarise the sample statistics, the decentralised fusion solution using Parzen representations yields a more accurate result.

Resource-aware distributed particle filtering for cluster-based object tracking in wireless camera networks

“…In addition to the referred distributed particle filters for camera networks, there are several distributed particle filtering approaches [9,[33][34][35][36] for wireless sensor networks.…”

“…For distributed particle filtering using multiple cameras, each camera computes its object probability locally, and the global object probability is obtained by fusing the local object probabilities of the camera nodes. In this case, however, the discrete nature of the method requires particle sample synchronization in computing the joint (global) probability [9,33,35,36]. Note that we use the term synchronization to describe the fact that all local particle sets have the same support points as defined in [9].…”

“…Hence, in many cases, it is preferable to encode the probability representation into a continuous form that is independent of the number of particles and does not require particle synchronization. Gaussian mixture model (GMM) [10,11,[40][41][42][43][44][45] and Parzen [33,[35][36][37] based probability conversion methods were proposed for converting particle probabilities into continuous forms. In this section, we show the derivation of a distributed particle filtering framework using the aforementioned probability conversion approaches, which will be utilized for the proposed tracking system.…”

Formation control driven by cooperative object tracking

“…The field of cooperative target tracking has gained a lot of attention in the recent years [14][15][16]. Many efficient solutions such as decentralized PF for multiple target tracking [15] and global position sharing based on non-egocentric tracking of objects [16] have been proposed.…”

“…Many efficient solutions such as decentralized PF for multiple target tracking [15] and global position sharing based on non-egocentric tracking of objects [16] have been proposed. These solutions focus more on compacting the data shared for communication bandwidth reduction and to overcome the problem of target occlusion.…”