A highly-scalable Bayesian approach to the problem of performing multi-source data fusion and target tracking in decentralized sensor networks is presented. Previous applications of decentralized data fusion have generally been restricted to uni-modal/uni-source sensor networks using Gaussian based approaches, such as the Kalman or information filter. However, with recent interest to employ complex, multi-modal/multi-source sensors which potentially exhibit observation and/or process non-linearities along with non-Gaussian distributions, the need to develop a more generalized and scalable method of decentralized data fusion using particle filters is required. The probabilistic approach featured in this work provides the ability to seamlessly integrate and efficiently fuse multi-source sensor data in the absence of any linearity and/or normality constraints. The architecture is fully decentralized and provides a methodology that scales extremely well to any growth in the number of targets or region of coverage. This multi-source data fusion architecture is capable of providing high-precision tracking performance in complex, non-linear/non-Gaussian operating environments. In addition, the architecture provides an unprecedented scaling capability for decentralized sensor networks as compared to similar architectures which communicate information using particle data, Gaussian mixture models, or Parzen density estimators.

No abstract

One approach to detecting swimmers with active sonar is to deploy a larger number of relatively simple nodes and network them together. Instead of using a complex multi-element phased array with electronic beamforming, this system uses air-backed parabolic reflectors, each with an omni-directional transducer. To avoid performance degrading acoustic interactions, the available operating frequency band is managed at the channel level. The physical configuration of this sonar system presents challenges for tracking through and across beams, and between nodes. Swimmers can exhibit low target strength and the acoustic clutter fields can be dense and highly dynamic. To detect and track swimmers in this environment, we employ a windowed Hough-transform (HT) tracker at the beam level. The HT has received wide use for track initiation. However, because of the processing gain required to continually track a weak target in such a significant clutter field, the HT is used in this case to maintain as well as to initiate tracks. The single- and multi-node tracker is a Kalman-based, multi-object tracker capable of tracking any number of stationary, constant-velocity and/or maneuvering targets. The multi-node tracker receives fused observations as inputs and creates accurate estimates based on the target position, velocity, and acceleration.

This paper presents a robust approach to the problem of localizing multiple emitters in cluttered environments using a network of low-cost sensors. A pseudo maximum likelihood (PML) function, whose peaks strongly indicate the number and locations of the true emitters, is generated from available sensor measurements. Localization is performed by first iteratively associating the individual sensor measurements to the PML peaks and then generating true maximum likelihood (ML) estimates of the emitter locations based on the associated measurements. These emitter location estimates are fed to an EKF-based, decentralized architecture for initiating, sustaining and sharing local track activity throughout the sensor network. The iterative PML approach is shown to be efficient with computational complexity that is linear in the number of measurements. Additionally, simulation results illustrate the ability of the proposed approach to generate accurate emitter location estimates in cluttered environments for generating and sustaining all track activity in a fully decentralized sensor network.

No abstract