Research On Some Key Issues Of Location Awareness Based Mobile Information Service

With the rapid development of Information&Communication Technology (ICT) recently, the fixed network-based communication pattern has gradually been replaced by wireless network-based one. Meanwhile, spatial information technology, especially the integration of Web GIS,GPS,RS,VR and Context Awareness,CSCW, has powerfully promoted socialization and popularity of the mobile information service. Being driven by technological progress and market needs, the Location Based Service (LBS) develops rapidly and plays more and more important role in various applications.LBS is moving oriented, so it is a mobile information service, which can get accessed information by virtue of location of mobile equipments through the mobile network. In fact, this application is a collection of internet, GIS/spatial database and a new generation of information and communication technology. It makes spatio-temporal locations of objects as index of related information and shields the gap between physical world and virtual world. Besides, its characters, such as transparency of the computation, seamlessness of the moving, generality of the information access and context-aware intellectuality, has made users get access to any location based information by any device and any network at any time and realized the people-centered information service rather than the machine-centered operation just as desktop pattern does.During these applications based on users and spatio-temporal location of moving objects, applying systems can continuously obtain the information of spatio-temporal location of users and moving objects through kinds of sensors. Meanwhile, the information will be stored in Moving Object Database (MOD) in the form of spatio-temporal trajectories. A large number of valuable characteristic information of users or moving objects, such as moving patterns and favored patterns, is implied in spatio-temporal trajectories, which reflects moving behaviors of moving objects in the MOD under three dimensional time space and is also the basis of applications mentioned above.The location aware-based mobile information service supplies relevant information according to spatio-temporal locations of moving objects and typical application is such as the information recommendation. The basis of it is also the spatio-temporal locations of moving objects. In a word, its key problem is to obtain, manage and mine the historical, present and future locations of moving objects. However, spatio-temporal locations of moving objects change continuously (Relative rest is a particular case of motion), so for location aware moving information service the key is transformed to the acquisition, management and mining for moving object trajectories.At present, the researches in location-aware fields mainly focus on the inquiry for historical and future locations implied in spatio-temporal trajectories, which doesn't do well in studying for moving behavior preference and characteristic patterns implied in trajectories and then querying future locations and judging moving intensions for users or moving objects. Affected by it, moving information service comes to make the present location of users to represent moving intensions of them, which is obviously unreasonable. First of all, different users may not have the same intensions even in the same spatio-temporal locations, so it cannot guarantee accuracy to recommend information according to users' present locations. Besides, even if present locations can represent users'moving intensions to some extent, it will still cause hysteresis to do so considering the time cost of system computing, which is especially obvious to users who move fast.In daily life, we know that the need for information is not only confined to historical and present information but impending incidents. However, present location aware-based moving information service cannot effectively mine common characters from historical data of users' movement, which cause low accuracy and low timeliness for future location awareness and is not enough intelligent to judge users' moving intensions. All of these cause serious influences on the applications of moving information serviceConsidering the situation that moving plans cannot be obtained previously, the historical spatio-temporal trajectory information is the only basis to judge if the future locations of moving objects are effective. Thus, it will be the key technique in location aware-based moving information service and also important points in the thesis to research how to get common characters from historical trajectory information of moving objects and discover biased patterns, according to which farther future spatio-temporal locations of moving objects will be inferred.Based on repeatability of moving behaviors for moving objects, the paper generates common moving preference patterns implied in historical spatio-temporal trajectories of moving objects. Then present trajectories of moving objects are compared with different moving preference patterns and the nearest one is found, according to which we can discover future father spatio-temporal locations of moving objects. It can be viewed the basis of the mobile information service. The paper does research on key problems from four aspects for location aware above which are respectively the schema of location-aware computing, model of location context, algorithm of location context organization and algorithm of future location context generation.1) The schema of location-aware computing. This paper proposes the characteristics of location-aware computing in comparison to comprehensive general context-aware computing. On that basis, a special location-aware computing conceptual model build on comprehensive general context-aware conceptual model is put forward. In the paper, it introduces the "extrospection-introspection" learning process in modern psychology in location-aware computing and proposes the "extrospection-introspection" context-aware computing process based on extrospection perception and introspection perception. Extrospection perception is to perceive available information related with spatio-temporal position in the external environment and then to proceed with deduce, decision and computation automatically, which can decrease manual participation to a great extent and then achieve transparent interaction and precise service. Introspection perception is to monitor the whole location-aware computing process and check context information process and reasoning method of the context-aware system itself. Then it can find problems from failures and low efficiency and further correct computing process and improve efficiency automatically. So extrospection and introspection supplement each other in location-aware computing.According to the "extrospection-introspection" learning process in modern psychology, it proposes a corresponding location-aware system model in the paper. The whole model is composed of lower context collection, context organization, high-level context application, application program interface, context storage and management and safety and privacy, in which context organization is the key and modeling, filtering, deducing, integration and introspection consist of main content in context information evolution. On the one hand, it can get high-level context information needed by each application by modeling, filtering, deducing and integrating lower context information and then extract moving preference model through moving historical information and further deduce context information in the future through current moving behavior, which is the basic and key function. On the other hand, it can monitor results of high-level context information and then judge, explain and correct failures and further feed back them to deduce and integration parts so that precision of the context organization is improved.2) Location context modeling. Space-time location context modeling is always one of the key problems in location-aware computing. Combined with the ontology-based model and spatio-temporal data model, the paper proposes and designs hierarchical model for location context which is consisted of two-layered structures. The location context model is divided into two layers which are respectively high-level Context Ontology Model and low-level Spatio-Temporal Data Model. Two layers are connected by query interface of model translator to solve the problem of separating the query and the reasoning in location context.The high-level Context Ontology Model is mainly to solve problems of reasoning and sharing and it is described by OWL. The key location context entity is defined as object, location and activity and expressed by the triples of RDF<Subject, Predicate, Value>. It describes physical entities or logical entities and their states in the moving environment to realize the sharing and precision of location context based on which the model focuses on discussing relations among key context entities, context state transition and life circle management of the location context.The low-level Spatio-Temporal Data Model is mainly to solve problems of storage and query. The paper discusses the model principle and information updating strategies and also defines the basic concept and designs the overall view of the spatio-temporal data model. As for the dynamic attribute of spatio-temporal location information, the paper analyses possible states and transitions of the location context and proposes the life circle management algorithm of location context information. As for present shortages of database of moving objects, the paper proposes the improved2-dimentional spatio-temporal data model of moving objects EBMOST and designs the overall structure of the spatio-temporal data model. As for the shortage that location updating strategies of databases of moving objects are just suitable for the point-like objects, the paper proposes a new location updating strategy for specific space-time. When the moving object is in a specific range where the environmental object is, MOD will view that the moving object has been in the location information of updated MOD. The location updating strategy can more effectively reflect the complex semantics and topological relations between moving objects and the environment. Besides, it to some extent compensates the shortage that the MOD cannot express face like objects after entity objects is abstracted zero dimension.3) Related algorithms of location context organization-moving preference pattern extraction algorithm. Aim at the location context computing in the future, the paper first put forward a trajectory interpolation algorithm according to the contradiction between continuity and discrete ways of location updating. The algorithm can shield the interference for the distance between trajectories caused by different locations updating strategies and sampling information granularity. And then, the paper introduces the Hausdorff distance and proposes the distance algorithm between spatio-temporal trajectories named IMHD-ST which measures differences between trajectories from temporal dimension and spatial dimension respectively and solves the problem of the basic distance measurement in the moving context organization. The paper then discusses semantics of the moving preference pattern. Based on these, the paper puts forward the moving preference pattern extraction algorithm named PPCG by adopting the mentioned distance algorithm between trajectories and density-based clustering algorithms. In the algorithm, it divides the moving preference pattern extraction into four steps which are partition, prune, clustering and generation. The partition is to merge similar straight sub trajectories to a straight trajectory so that the time costing can be decreased on the premise of limited loss of precision. The prune is to get rid of the noise in straight sub trajectories and keep similar ones. The classification is to reduce remained straight sub trajectories into complete trajectories and carry out clustering. The generation is finally to generate the moving preference trajectory in three-dimensional space-time. Thus, the algorithm accomplishes the extraction from discrete spatio-temporal location updating points of moving objects to continuous moving preference pattern.What's more, in the partition step, it proposes MHD-S-based fixed threshold partitioning algorithm named MHDS-Partition. Without the increase of time complexity, the algorithm can effectively find characteristic location points among temporal trajectories and improve precision and simplicity of trajectory partitioning. In the generation step, it proposes moving preference trajectory generation algorithm based on scanning circles and vertical scanning lines. Besides, it gives the changing algorithm from static and offline process pattern to dynamic and online process pattern.4) Related algorithms of moving context organization-future location context computing algorithm. The future location context computing is the most important algorithm in future location context prediction. By comparing and analyzing existing methods and then combing the particularity of spatio-temporal location context information, the paper proposes a performance complementary algorithm based on integrated naive Bayes Classifier and linear regression. So the algorithm not only possesses the precision of the naive Bayes Classifier but also velocity of the linear regression to meet the needs of particularity in spatio-temporal location context information and real time in the moving information service.On the algorithm design, the paper first proposes the Boosted Dynamic Naive Bayes Classifier for MTP (BDNBCM). The Naive Bayes Classifier is famous for its high efficiency and integrated methods can optimize classifying performance to a great extent, so the two are combined to process dynamic changes and incomplete spatio-temporal location context information in dynamic preference trajectory classifying model, which can improve prediction precision of future spatio-temporal location. Besides, in the paper, it also gives chosen thoughts and related algorithms of the quantity k of the dynamic classifier and makes the computing cost least on the premise of meeting the needs of classifying requests. What's more, the paper shows the condition density calculation method of the above classifying model parameters. Owning to the particularity of spatio-temporal location information, it is difficult to directly estimate related parameters of moving patterns or moving trajectories. So in the paper, it makes continuous moving preference trajectories as media and dynamically obtains corresponding parameters estimation of each location updating point in classified trajectories based on the interpolation algorithm.Besides, the paper does researches on regression-based future spatio-temporal location algorithms and proposes the regression algorithm of future location context. Based on future moving plans and current moving states obtained through classification-based calculation, the algorithm translates regression calculation into the query in the index of the time, which changes the method in previous references that moving patterns are fitted into a certain function which makes the time as the variable. Owning to the restriction of future moving plans and correction of current moving states, the algorithm performs better on prediction precision compared with ones without restriction and correction.The paper aims to realize exact computation of the high-level location context information such as the future spatio-temporal locations for users and does a serious of in-depth studies on it. What's more, the paper proposes efficient and practical solutions for several related questions which are involved in the computation.The innovations of the paper are as follows:firstly, the psychological concept of "extrospection-introspection" is introduced and specified computation concept model of location awareness is built and then corresponding system model is proposed. Secondly, the paper puts forward and builds two-layers hierarchical model of the location context, which possesses advantages of ontology model and spatio-temporal data model; thirdly, the algorithm of Hausdorff-based three dimensional spatio-temporal (two dimensional plane and one dimensional time) trajectory similarity is proposed, in which two dimension is plane and one; Fourthly, it puts forward the mobile preference pattern extraction method which is based on density clustering; fifth, it proposes an algorithm of future location context which combines integrated Naive Bayes Classifier and linear regression.The main contribution of the paper is that it puts forward a systematic spatio-temporal location prediction method and related algorithms, which can more accurately and rapidly calculate the high-level position information such as future spatio-temporal location according to historical positions under the existing technical conditions.The paper makes detailed performances analysis for proposed algorithms and adopts simulated data sets or actual data sets to do detailed experiments at different parameters. A number of research results have been published on international meetings or national core journals and retrieved by El or SCI-E, which includes ICCSE2010, APCIP2010,《Journal of Computational information Science》 and 《China Communications》. Experiments show that algorithms in the paper can effectively solve related problems and possess complete functions and good performances compared with existing ones. Therefore, researches in the paper can provide great theories and applications for many key technologies on location-aware mobile information service.The real data sets come from data sets of global (including the Atlantic, the eastern Pacific, the western Pacific, the southern Pacific, the southern Indian Ocean and the northern Indian Ocean) historical hurricane trajectories which are provided by an American website, unisys.