Research On Entity State Matching Estimation Methods Towards Search Service In The Internet Of Things

With the wide deployments of a variety of different forms and versatile sensing devices in the physical world, an increasing number of physical entities have released and shared their state information on the Internet through the sensor. Facing more and more sensors and entity status data in the physical world, only the use of efficient search technologies can reflect the vitality of these data and achieve the value of sensing resources in the Internet of Things (IoT).In the IoT, the search objects are relatively wider, the sensor resources are limited,and the search contents are dynamically changing. The technical solution that releases search command, matches web content, processes web pages, and returns search results adopted by the traditional Internet search cannot effectively solve the problems faced by IoT search. The current cutting-edge research results point out that the efficiency of IoT search can be improved by predicting the match state of physical entity. However,there are mass of sensors whose functions and strengths are diverse and uneven. The existing research achievements have not taken into account the constraints of search scenes, the needs of search patterns and the resource characteristics of search platforms.Hence, they fail to propose more appropriate and effective match prediction methods.In view of existing problems, this thesis proposed a distributed entity state match assessing method,provided a match forecasting method for entity search in the IoT resource-limited scenario, and designed a low-overhead entity state match prediction method for Internet of things search.(1) Due to the reason that the existing achievements do not support the search mode which can return all the search results and ignore the search needs for entities that will meet the specific content in the future time. Moreover, the performances of existing distributed entity state match forecasting methods are very limited. Hence, a mode adaptive search scheme that can return all match results and meet the requirement to future search was proposed. A time-dependent distributed entity state prediction method was designed to accurately estimate the entity future state, so as to solve the time-independent and low prediction precision problems existing in the current research. An entity matching estimation and verification method was provided for solving the large communication overhead cost problem caused by the traversal access process.Compared with the existing methods, the recall rate, precision and communication overhead performances can be respectively improved more than 34%, 13% and 20%by the proposed methods.(2) Aiming at the problem of the lack of efficient entity search matching estimation method and the fact that the research on entity search mechanism towards the resource-restricted IoT search scene is insufficient, an entity search system for resource-constrained IoT was proposed in this thesis. A search solution applied to resource-constrained platforms was designed. A high-precision entity state prediction method applied to resource-constrained scene was presented,including the equal-interval and during the period prediction method. And then a low-cost entity preferentially verifying method was proposed based on the predicted entity state to reduce the communication overhead of verification process. Compared with the existing methods, the proposed methods can effectively reduce the resource requirements to the sensor, improve the prediction accuracy of the physical entity state by more than 5%, and enhance large than 15% of the communication overhead performance of search system.(3) Contraposing the large consumption problem of search matching estimation methods in the existing research, this thesis put forward an advanced low-cost entity state match prediction method which fully considered the resource characteristics of the sensor and the gateway platform. Furthermore, an efficient search strategy was proposed. The proposed match prediction method contained a lightweight data fitting method, which aimed to fit the data at the cost of low computational overhead and solve the consumption problem existing in the current entity state data reporting methods.And it also includes a highly sparse entity multi-step state prediction method, which was presented to achieve high-precision prediction performance of entity multi-step state with lower computation and storage cost, so as to solve the large resource consumption problem caused by the search matching method adopted by the gateway.Results demonstrate that the proposed methods can lower the transmission energy consumption by more than 61%, and reduce the calculation and storage overhead by more than 65% at the expense of a certain communication overhead performance.