Wireless Localization And Tracking Based On Bayesian Estimation

Wireless localization and tracking is of critical importance in lots of location-dependent applications, such as ubiquitous computing and communication, intelligent sensing, patients or elderly monitoring, intrusion detection, battlefield surveillance, etc.. However, due to the complex environment, shadowing of the obstacles, and the deficiency of the algorithms, how to realize robust accurate localization and tracking under complex circumstance becomes a challenging and promising problem. This dissertation focuses mainly on the study of the schemes to realize robust wireless localization and tracking based on the Bayesian estimation theory. The main contributions of the dissertation are summarized as follows:Firstly, to overcome the degeneration problem of the traditional particle filter (PF) algorithm, an improved PF algorithm is proposed, which makes full use of the latest observation in constructing the proposal distribution. The quality prediction function is proposed to select the promising particles, and the centroid shift vector is defined to lead the particles move towards the optimal proposal distribution. This work provides the foundation for the subsequent research works.And then, the dissertation researches the schemes to realize robust wireless localization and tracking in mobile wireless sensor networks (WSNs), indoor localization, and device free localization. The main contributions are summarized as follows:1. To improve the Localization and tracking performance in mobile WSNs, an enhanced Monte Carlo algorithm is proposed. The algorithm adopts the controlled flood method to improve the using efficiency of the anchor nodes, the genetic cross operation to accelerate the sampling process, and the interpolation operation to predict the velocity and angle of the target. Meanwhile, the signal character sequence is defined and adopted to translate the signal of the physical layer to the abstract layer, and to deal with the update of the particle weights efficiently.2. To capture the dynamic signal propagating character in the complex indoor environment, a novel differential radio map is proposed and a localization algorithm based the differential radio map is designed, where the location is estimated according to the difference of received signal strength (RSS) instead of RSS itself. Meanwhile, to overcome the non-line-of-sight (NLOS) effect incurred by the reflection and diffraction, a novel joint states estimation localization algorithm is proposed, which adopts a Markov model for NLOS state estimation and a particle filter for location state estimation. The feasible region of the particles is built based on the NLOS and line-of-sight measurements, which utilizes the positive effect of the NLOS measurements while restraining their negative effect.3. An approach for tracking targets in wireless networks without the need of equipping the target with a wireless device is proposed based on the principle of radio tomography. According to the shadowing effect caused by the target, a dynamic statistical model for relating the change of the RSS between the node pairs to the spatial locations of the targets is presented. The schemes of utilizing the statistical model to build the observation likelihood function of the target and realizing the target localization and tracking under the particle filter framework are investigated. Meanwhile, the problem is formulated as a sparse signal recovery problem, and a novel Bayesian greedy matching pursuit (BGMP) algorithm is proposed to tackle the signal recovery problem even from a small set of measurements. The BGMP utilizes the past estimations to build the enumeration region, so as to speed up the algorithm and improve its recover performance simultaneously.Finally, to prolong the lifetime of the wireless nodes, a radio-trigged wake-up circuit is proposed to control the activation and shutting down of the node, and thus eliminates the energy wasting wake-up periods. A wireless ranging sensor node based on the chirp-spread-spectrum is developed, which has the ability of realizing accurate ranging and does no need the time synchronization between the nodes.The above works enrich the theory of wireless localization and tracking, and make some valuable exploration on the applications of the theory in the related field.