Research On Key Cooperative Localization Technologies In Wireless Ad Hoc Networks

With the development of wireless networks,there is an increasing demand for accurate and reliable position information.Traditional localization techniques(e.g.,GNSS)are now faced with striking challenges because the environment is becoming more and more complex.The localization system based on wireless ad hoc networks is a promising solution for accurate localization in harsh environments due to its flexibility and ease of deployment.By introducing the measurements between user nodes,cooperative localization(CL)technique can improve the localization accuracy and reliability of the nodes and meanwhile expand the service area,which has attracted significant attention and research interests.Despite its many advantages,CL is also faced with some problems.For example,nonlinear measurement model makes it difficult to effectively fuse the measurements,especially in sparse and mobile ad hoc networks;in dense networks,CL will dramatically increase the communication overhead and energy consumption.Besides,using inaccurate measurements from some unreliable neighbors may result in error propagation throughout the whole network;in mobile networks,such as Internet of Vehicles and unmanned aerial vehicle networks where the node positions and measurements are fast changing,high position information update rate and small information transfer delay are required to meet the demands of real-time position estimations,which makes the efficient fusion of the measurements and optimization of resource allocation become more challenging.Focusing on several CL problems in wireless ad hoc networks,the thesis will investigate reliable localization methods in sparse and mobile environments,and neighbor selection and resource allocation algorithms in dense networks.Meanwhile,the thesis will analyze the impact of measurement timeliness on the localzaition performance of mobile networks.The content organization and main conclusions of the thesis are listed as follows:Focusing on sparse and mobile CL environments,the thesis first investigates reliable localization methods.On the one hand,the thesis proposes a mobility constrained CL algotirhm which has an embedded double loop structure.In the inner loop,sum-product algorithm is utilized to fuse range measurements and the results are used as the observations of the outer loop algorithm.In the outer loop,an algorihtm based on the extended Kalman filter framework is used to update position estimations of the nodes,and meanwhile velocity estimations of the nodes can be obtained.On the other hand,the thesis proposes an extended joint spatial and temporal cooperation(JSTC)model where the position estimations of the nodes at the current time slot can be obtained by combining the range measurements of several recent time slots subject to the mobility constraints.Based on the extended JSTC model and maximum likelihood criterion,both centralized and distributed problems are formulated,and then solved by the sequential quadratic programming algorithm.The distributed problem takes the cooperation among multi-hop nodes and the information transfer delay into consideration.Besides,the performance limits of these two problems are analyzed by the fisher information matrix(FIM).Simulation results show that the mobility constrained algorithm is more accurate and reliable than the competing algorithms in sparse and mobile environments.The performance of the extended JSTC model is superior to the model which only considers a single time slot measurements.Second,focusing on the dense networks with limited resources(e.g.,bandwidth,time and energy,etc.),the thesis investigates the neighbor selection and resource allocation algorithm design.Specifically,based on the squared position error bound(SPEB)criterion,the thesis formulates two neighbor selection problems under implicit and explicit energy constraints,respectively,to balance the energy consumption of the network.In the implicit case,the thesis proposes to limit the number of times that each neighbor can be used in multiple consecutive time slots.While in the explicit case,the neighbor selection and power allocation among the possible neighbors are jointly optimized to further improve the localization performance.The resulting problems are difficult to solve due to the nonlinear objective functions and discrete optimization variables.The thesis proposes to transform them into more tractable forms by carefully introducing auxiliary variables and then employs the penalty dual decomposition(PDD)framework to solve the transformed problems.Simulation results show that the proposed neighbor selection scheme can significantly reduce the resource consumption with minor localization performance loss,and meanwhile the energy consumptions of the neighbors can be well-balanced.The proposed PDD based algorithms can significantly outperform benchmark algorithms.In particular,for the problem under implicit energy constraints,the performance of the PDD based algorithm is near-optimal.Finally,focusing on the CL system based on time division multiple access protocol,the thesis investigates the localization performance of mobile nodes which is affected by the measurement timeliness,and designs the broadcasting node selection and time slot allocation algorithms.The thesis first employs the age of information concept to model the timeliness of the measurements,and then by combining the mobility model of the nodes,the performance limits of the nodes in one superframe and multiple consecutive superframes are respectively analyzed by the FIM.In addition,the thesis also provides the performance limits analysis of the nodes in the stable state.In order to alleviate the negative effects caused by the measurement timeliness and increase the position information update rate,a joint broadcasting node selection and time slot allocation problem is formulated based on the SPEB criterion.The resulting problem has a special structure and the thesis divides it into a broadcasting node selection problem and a time slot allocation problem,and then a PDD based algorithm and a greedy algorithm are developed to solve these two problems,respectively.Simulation results show that the PDD based broadcasting node selection algorithm outperforms the random selection algorithm.The greedy based time slot allocation algorithm achieves similar performance to the random allocation algorithm with much lower complexity.