Research On Adaptable Positioning In Complicated Environment

Navigation and localization are becoming more and more important in industry,military,and our daily lives.The widely used GNSS only works well in open sky outdoor environment,where at least four satellites are visible.In the other environments,such as canyon,dense foliage,underground,urban canyon and deep indoor,GNSS fails because lack of enough satellites for positioning.The weakness of GNSS begins to restrict the development of navigation industry.This thesis aims to address the challenge that positioning in GNSS challenged environments,I propose a adaptive indoor and outdoor navigation and localization algorithm.The main research work and innovations are as follows:1.I propose a adaptive indoor and outdoor navigation and localization framework.We might be in all kinds of environments in our daily lives,including forest,canyon,crowded avenues,underground,deep indoors and so on.As we know,no single positioning technology is capable of performing well in different kinds of environments.However,there is accurate enough positioning technique for any specific environment.For example,GNSS performs well in open sky environment,while Shadow Matching is enough for urban canyon environment,and Wi-Fi fingerprint positioning is suitable for indoor environment.This implies us that we can integrate these positioning techniques for adaptive indoor and out navigation and localization.Selecting different positioning techniques for different environment is an alternative method.I propose an adaptive positioning framework for complicated environment.The framework consists of two processes-data collection and context sense,algorithm selection and positioning.I first record signal strength in the environment,and determine the environment where we located in.According to the type of environment,I select different algorithm for positioning.This framework is robust and extendable.2.I propose to use the GSM signal strength to detect the context.For this purpose,I propose to use the GSM cellular base stations' signal strength to detect the environment.The basic idea is simple: the propagation of radio signal is affected by the environment.Different environment results in different signal strength characters.By identifying the signal strength's characters,I can determine the user's environment.The context sense algorithm contains three phases: Data Input,Training,and Testing.In the Data Input process,neighboring cellular base stations' signal strength is recorded,and features are extracted.The data is classified in the Training phase.The classifier is applied to detect the user's current environment in the Testing phase.I investigate a wide range of Machine Learning algorithm for classification,including Decision Tree(DT),Random Forest(RF),Support Vector Machine(SVM),K Nearest Neighbor(KNN),Logistic Regression(LR),Naive Bayesian(NB),and Artificial Neural Network(ANN).3.For the light indoor environment,I propose to apply the hybrid cooperative positioning for localization.Because of exchanging information between neighboring users,the number of nodes,including satellites in view and localized neighboring users,for positioning increases.As a result,the receiver may fail to estimate their position in real-time.I propose an improved quosi-optimal algorithm to select both satellites in view and localized neighboring nodes for positioning.Hybrid cooperative positioning is not suitable for smart phones.Because ranging between phones in complex indoor environment is not accurate enough.I propose a range-free hybrid cooperative positioning algorithm.The proposed algorithm is fully distributed,and suitable for smart phones.The proposed algorithm is capable of extends the positioning coverage,and shorten the average positioning time.4.I propose to use fingerprint localization algorithm for the deep indoor environment,where no satellite available for positioning.Fingerprint localization is the most widely used algorithm in deep indoor environment.But building the database for fingerprinting is labor intensive.I first we propose Local Gaussian Process to create a virtual database using the training signal database.The virtual database,contains fixed number of reference points,is used for positioning.And the training database,created by user crowdsourcing,is used for updating the virtual database.Fingerprint based localization is also hampered by the variations of the received signal strength(RSS)due to e.g.impediments in the channel,decreasing the positioning accuracy.In order to improve the accuracy,we integrate Pedestrian Dead Reckoning(PDR)with Wi-Fi fingerprinting: the movement distance and walking direction,obtained with the PDR algorithm,are combined with the K-Weighted Nearest Neighbor(KWNN)algorithm to assist in selecting reference points(RPs)closer to the actual position.Seamless indoor and outdoor Navigation And Localization is full of challenges,and is of great importance both in military and civil field,which implying our research has great theoretical and practical significance.This thesis is the first step in this subject,and we will continue my research in the future.