Research On Modeling And Optimization Approaches For Device-Free Localization

Device-Free Localization(DFL)is an emerging wireless localization technique to estimate the location of the target that does not attach any electronic devices.It has been widely used in many fields,such as intrusion detection,emergency rescue,and aged health care.DFL is usually implemented by analyzing the changes of the received signal strength(RSS)measurements of the wireless links.However,due to the uncertainty factors,such as dynamic environment,non-line of sight and multipath propagation of wireless commumication,high-accuracy DFL with strong anti-interference capability is remaining as a challenging open research problem to address.Building effective signal propagation model is an effective way to improve the accuracy of DFL.This dissertation gives a comprehensive study for this problem and its corresponding optimization approaches.The main works are as follows.Firstly,the factors that affect the changes of the RSS measurements of the wireless links are discussed,and a joint link detection approach is proposed to select affected links collaboratively and reject the outlier links.Secondly,the optimization approaches for DFL are studies based on the geometrical link detection models.A nonlinear optimization approach for DFL is proposed by using the line detection model and convex optimization.In order to improve the accuracy of DFL,a convex feasibility problem is formulated for DFL based on the elliptical weight model,and parallel projections onto convex sets algorithm are proposed as a solution.The experimental results demonstrate that the proposed approaches are robust to the fluctuation of the wireless signals with superior localization accuracy.Thirdly,to deal with the worse localization accuracy under complex environment,the main factors affecting the changes of the RSS measurements of the wireless links are analyzed theoretically and verified by experiments.A Gaussian process based radio frequency signal propagation model is proposed and its performance is verified by experiments.Experiment results demonstrate that the Gaussian process based radio frequency signal propagation model is effective to predict the RSS measurements of the wireless links.Finally,based on the Gaussian process radio frequency signal propagation model and the joint link detection method,DFL is reformulated as a maximum likelihood probability optimization problem,and a modified Fruit Fly Optimization approach is proposed to deal with the problem.In addition,Gaussian process radio frequency signal propagation model based particle filter approach is proposed for device-free tracking problem in which the particles are predicted and updated based on the target motion model.Outdoor and indoor experimental test-beds are set up to evaluate the performance of the proposed approaches,and the results demonstrate that the proposed approaches are superior to the existing approaches.