On The Statistical Bayesian Analysis-based Wireless Network Localization And Tracking Technique

Cooperative localization and mobile tracking technology in wireless ad-hoc networks attracts more and more research attentions, with the rapid advances in 5G communica-tions, vehicular ad-hoc networks and internet of things.The wireless localization and tracking technology defines an efficient way that peo-ple search, locate and navigate the points of interest inside a local area. In some closed areas, such as a shopping-mall centre, the performance of global positioning system (G-PS) is limited. Hence, the localization methodology defined in a specified wireless sensor network (WSN) or the existing wireless local area network (WLAN) is promising. How-ever, in practical scenarios, there are a multitude of dependent factors could degrade the achieved localization and tracking performance, such as the inherent distortion of obser-vation methodology, the localization background disturbance, and the undesirable shift in localization network deployment.Firstly, the existing measurement methodology is a non-linear function of the rel-ative location, distance or angle, e.g., received signal strength (RSS), time of arrival(TOA), time difference of arrival (TDOA) and angle of arrival (AOA). The non-linearity of measurement function can not only result in a non-convex objective function (e.g.,the posteriori density), but also lead to the error accumulation effect, which can compli-cate or degrade the practical localization and tracking system. In order to mitigate the impact of non-linear measurement methodology, in this paper, some stochastic particle-based localization methods are proposed (e.g., variational particle filtering-based mobile tracking algorithm and variational inference-based network localization method). In ad-dition, a stochastic particle-assisted search algorithm is proposed to solve the non-convex optimization problem in cooperative localization.Secondly,due to the different surroundings and different degrees of thermal noises,the measurement accuracies at different reference nodes are commonly different (spatial-field randomness). In particular, due to the movement of mobile target, its reference nodes possibly change over time and the true values of their measurement accuracies are hard to capture. Hence the measurement accuracies of reference nodes are time-varying as well(temporal-field randomness). Namely, the measurement accuracy is nondeterministic and unknown. In light of this, the spatial-temporal-field randomness of measurement accuracy is proposed in this paper, and a Wishart density is employed to model its uncertainty. On this basis, the variational inference-based positioning and mobile tracking algorithms are proposed. The impact of uncertain measurement accuracy on localization and tracking performance is also disclosed in theory.Thirdly, due to the inevitable error in the initial location acquisition of network nodes, and due to the possible location shift by some unexpected operation, the reference node locations are usually inaccurate. Particularly when the number of anchor nodes in the network is limited, the network nodes with inaccurate locations are commonly used to perform localization cooperation. Hence, the reference node location error is a crucial factor for localization performance. In this paper, this factor is considered in both system modeling and algorithm design of cooperative localization and mobile tracking scheme to gain as more performance as possible.Fourthly, the network localization suffers more from the reference node location er-ror and measurement accuracy uncertainty. Consider a distributed network localization,wherein almost all of network node locations are inaccurate and need to be improved through mutual localization cooperation.Given the network measurement data and ini-tial (inaccurate) location information a distributed cooperative localization of network nodes is proposed in this paper. The performance limits under the initial conditions given are also investigated to provide some insights. In addition, the concept of localiza-tion information propagation is proposed to reveal the essence of spatial-field localization cooperation and error propagation in network localization.Fifthly, for the mobile tacking, the movement manner of mobile target is diverse,wherein the underlying information is two-fold: the regularity (e.g. the frequent mode that can be modeled) and the randomness (e.g., the unexpected shift that is difficult to predict). In order to more generally characterize the target mobility and measurement data, to more deeply capture the underlying information,a multilayer dynamic Bayesian network (MDBN) model is proposed in this paper. Based on this MDBN model, a vari-ational Bayesian filtering-based joint mobile tracking and velocity prediction scheme is proposed as well. The proposed joint scheme can be applied to track both a directional-moving target and a random-moving target. At the same time the underlying velocity information can also be identified to further improve the mobile tracking performance.In fact, the tracking of mobile target could be regarded as the localization cooperation in the temporal-field.Hence based on the concept of localization information proposed in this paper the principle of temporal-field localization information propagation is also investigated in theory. In addition, the convergence condition of mobile tracking error and the tracking performance limits are revealed as well. In particular, it is disclosed that, the essence of temporal-field localization cooperation is the propagation of localization information in temporal-field.In brief, different from the GPS localization, this paper focuses on the cooperative localization and tracking issue in wireless ad-hoc networks and some distortion factors(e.g., reference node location error, indeterministic measurement accuracy, diverse target mobility, nonlinear measurement methodology, ete.) are highlighted to investigate and mitigate their impacts to reap more gains as possible. Focusing on these distortion factors,the cooperative localization and mobile tracking are investigated in this paper, from sys-tem model, algorithm design & performance analysis. As a result, some robust algorithms are proposed, the spatial-field and temporal-field localization cooperation principles are revealed,and finally the performance limits over those dependent factors are quantified.