Research On Internet Of Things Resource Management Platform And Service Provision Framework

Most of the application-modes of existing IoT systems are "silo" applications, and IoT is still stay in the phase of Intranet of Things. This mode is characterized by "one device-one application." In some cases, even special networks are being built for single applications. So, the resources and data are locked in each application "silo", and cannot be reused and shared. Every application "silos" are isolated with each other, the information of each application cannot be interconnected, therefore they form many information isolated islands. This mode wastes large amounts of resources, and hinders the further intelligence fusion of cross-system information. Moreover, it cannot support future more pervasive and large-scale application solutions. This paper focuses on providing an infrastructure to open up or even break the current application silos and facilitate the bridge from Intranet of Things to Internet of Things.This paper proposes a comprehensive IoT service provisioning framework which provides the infrastructures for the customizable openness and sharing of data and resources under the premise of ensuring the real-time and reliability of applications. The resource management platform is the key component of framework. It aims to open resources’ capabilities and enable the reusing and sharing of cross-system resources, meanwhile support the interaction between systems. For the functional aspect, it provides an infrastructure for heterogeneous resource accessing and formally describing resources and entities, then provides their outputs to different applications by the way of well-understood and machine-processible format. By the validation of several practical projects, the performance and stability of platform is competent to the requirements of practical IoT applications.For the issue of resource heterogeneity, this paper mainly focuses on protocol heterogeneity and data model heterogeneity. The platform adapts the heterogeneity of device protocol by the layer of UDA (Unified Device Access) in which the protocol stack framework is the key component. Protocol stack frame is based on OSGi platform and employs the protocol stack design Pattern of Dependency-Inversion. It has two features:it supports the PnP (Plug and Play) of protocol, namely, the protocol can be installed, updated and uninstalled dynamically; it supports the self-adaption of protocols, namely, it can automatically choose protocol stack or assemble protocol stack to analyze unknown data packets without manual configuration. Besides, the protocols of protocol stack framework are contained in a protocol container in the form of OSGi bundles. This container provides a GUI management tool to maintain, manage and configure deployed protocols. By the evaluation of comprehensive experiments, the results indicate that the performance of UDA is competent to the time-critical applications.For the heterogeneity of data model, this paper proposes a multi-level and multi-dimensional information model base on semantic ontology. Heterogeneous resources are modeled and described uniformly by information model after they are accessed by UDA. The idea of multilevel is that underneath resource model and upper entity model are decoupled from each other, they are provided by different business roles. They can be linked by resource-entity binding, after that, resource can observe or control the properties of entity. Then, unmeaning sensing data can be interpreted to context data which have the dedicated meaning of specific applications. The entity model from one application can be linked to several resource models provided by different providers, which means that it enables the reusing and sharing of resources and data from the level of model. Besides, information model provides the mechanism of model linking and mapping. By this mechanism, the data from sensing layer or other systems can be provided to the upper applications by the syntax or semantic they can understand. Moreover, things have multi-dimensional features, each dimension should be described by different taxonomies and models. The idea of multi-dimensions describes IoT resources by the way of multi-dimensional resource model. Based on the multi-dimensional resource model, resource matching algorithms can accurately match resources according to their features from different dimensions. Besides, this paper proposes a domain template based modeling tool by which users can create, organize, and manage their information model. The information model proposed has been employed in several projects, such as Siemens Intelligent Transportation System and District Heating Control and Information System, etc.After large-scale resources are accessed into platform, their capabilities are exposed in the form of lightweight services. This type of service is named as real-world service. In consideration of the large-scale, dynamic and context-complex features of IoT services, this paper proposes the semantic similarity based functional service recommendation and the QoS-based service recommendation according to collaborative-Filtering. They recommend appropriate services to users automatically or semi-automatically according to the requirements of users and the context of monitored-entity (e.g. location, quantity type and environment information, etc). The algorithms focus on two aspects:they assist the users to select interested services from large-scale available services; and when the context of resource or entity changes, they find substitute resources which are confonn to the current context of entity, thus ensuring the continuity of real-world service. The results of comprehensive experiments indicate that the proposed service recommendation algorithms are suitable for IoT environment. They can ensure the performance of accuracy and recall within acceptable time.