C4F

Advanced multi-sensor systems, either using a fusion center or by a peer-to-peer (P2P) wireless network, are expected to combat the challenges that arise in object recognition and state estimation in harsh environments with poor or even no prior information, while bringing new challenges mainly related to data fusion and computational burden. Unlike the prevailing Markov-Bayes framework that is the basis of a large variety of stochastic filters and the approximate, we propose a clustering-based methodology for multi-object detection and estimation (MODE), named clustering for filtering (C4F), which abandons unrealistic assumptions with respect to the objects, background and sensors. Rather, based on cluster analysis of the input multi-sensor data, the C4F approach needs no prior knowledge about the latent objects (whether quantity or dynamics), can handle time-varying uncertainties regarding background and sensors such as noises, clutter and misdetection, and does so computationally fast. This offers an inherently robust and computationally efficient alternative to conventional Markov-Bayes filters for dealing with the scenario with little prior knowledge but rich observation data.

1 - Centralized implementation - multi-sensor C4F

The first present C4F solution relies on the cluster analysis of the observation data that are synchronously reported by multiple/massive sensors, we omitting the issues regarding the communication between sensors. Attached are the source code and the database for an illustrative C4F example implemented in a MODE scenario with NO prior information. The simulation is set up in the most common manner, as widespread in the literature. The results appeared on the following work

T.Li, J.M. Corchado, S. Sun and J. Bajo, Clustering for filtering: Multi-target Detection and Estimation Using Multiple/Massive Sensors, Information Sciences. Volumes 388–389, May 2017, Pages 172–190. @ ScienceDirect

See subpage (O2/C4F) for more details and the preprint of the paper. Matlab Codes available:

Codes: C4F_CentralizedExp

--------------------------------------------------------------------------------------

2 - Distributed C4F - aided by Flooding for measurement communication

Based on a distributed sensor network --- again, we consider very typical and common distributed sensor network setup --- we develop a flooding-then-clustering (FTC) approach which comprises two parts: (1) a distributed flooding scheme for iteratively sharing the measurements between sensors and (2) a clustering-for-filtering approach for target detection and position estimation from the local aggregated measurements.

The clustering implemented is a simplified version of the above centralized C4F approach, considering the realistic situation that the the severely limited computing and communication capability of the low-powered sensors.

We compare the model-free FTC approach with three types of distributed probability hypothesis density (PHD) filters that are ideally modeled with statistical knowledge about the target motion and the sensors. The three PHD filters include (1) multi-sensor Iterated-corrector PHD filter, (2) Arithmetic average PHD Filtering (based on either GM, SMC or hybrid), see: arxiv.org/abs/1711.10783 and arxiv.org/abs/1712.06128 and (3) Track-to-track fusion based on PHD filter (no fusion feedback)

A series of simulation studies using either linear or nonlinear sensors, have been presented to verify the effectiveness of the FTC approach. Matlab Codes are available at the following link corresponding to the three simulations to be presented in the paper:

T. Li ; J. M. Corchado ; J. Prieto, Convergence of distributed flooding and its application for distributed Bayesian filtering, IEEE Trans. Signal and Information Processing over Networks, vol.3, no.3, pp. 580 - 591. @ IEEE Xplore
T. Li, J.M. Corchado and H. Chen, Distributed Flooding-then-Clustering: A Lazy Networking Approach for Distributed Multiple Target Tracking, FUSION 2018, Cambridge, July 10-13 2018. @IEEE Xplore

codes: Distributed C4F_Flooding-then-clustering

3 - Working with the RFS filter - multi-sensor measurement clustering for filtering

Data-driven, learning methods do not conflict with the model-driven filters, at all!

This letter presents a novel multi-sensor probability hypothesis density (PHD) filter for multi-target tracking by means of multiple or even massive sensors that are linked by a fusion center or by a peer-to-peer network.

As the challenge we confront, little is known about the statistical properties of the sensors in terms of their measurement noise, clutter, target detection probability and even potential cross-correlation. Our approach converts the collection of the measurements of different sensors to a set of proxy, homologous measurements. These synthetic measurements overcome the problems of false and missing data and of unknown statistics and facilitate linear PHD updating that amounts to the standard PHD filtering with no false and missing data. Simulation has demonstrated the advantages and limitations of our approach in comparison to the cutting-edge multi-sensor/distributed PHD filters.

T. Li , J. Prieto ; H. Fan ; J. M. Corchado，A Robust Multi-Sensor PHD Filter Based on Multi-Sensor Measurement Clustering，IEEE Communications Letters, vol.22, no.10, pp. 2064 - 2067. @IEEE Xplore