| Internet of Things (IoT) has attracted ubiquitous attention in both industrial and academical regions as a great important domain. IoTs integrate heterogeneous network including wireless sensor network, telecommunication network, internet, etc., face at mobile and fixed terminals and cover various services like monitoring, transfering and computing, thus the management of IoT is extremely complicated.This thesis analyzes problems emerged from the four layers including sensor, network, data and application in the architecture, data, spectrum, heterogeneouse network and security domains and is expected to address these complex issues in a mutiple/cross-layer way. Aiming at some shortcomings of state of art, several novel or improved methods and algorithms are proposed to point at each problems. Fully consideraing the diversity and heterogeneity of IoT ecosystem, the proposed methods and algorithms are pervasive for various network protocols.The contributions of the thesis contains:1. A three-hierachy (node, cloud and local) interaction and configuration architecture is proposed, which guarantees each layer can process data and deploys programs or softwares in a smart and flexible way. Also, using Plug-and-Play sensor module, TI-CC2530, ARM11and Service, a test platform has been built as supporting the three-hierachy architecure.2. Combined sensor and data layer, as facing at the network conjection caused by data deluge, it is proposed that a granular-based sensor node access scheme which can control the scale of data transmission. Granular is utilized to select more representive data to transfer. The number of data attributes can be selected in1to N or2to N optionally. It is very flexible and non-character based. Additionally, a data-padding method is used since the power graph based granular computing is not suitable for incompleted data. Comparing with KNN imputation, the proposed method is more reasonable and rather suitable for large scale and online task for nearly no extra computation.3. In network layer, a Stackelberg game is modeled. Two parties’behaviors are abstracted as maximizing both benefits. With cross-layered parameter design, the model is systematized using satisfactory function. Comparing with other Stackelberg game, the modeled game is more reasonable in cooperation mode. Besides, the complexity is lower since the primary user does not need to negotiate price with every second users. 4. Combined network and application layers, from heterogeneous network aspect, this thesis discusses two kinds of user preference. For specific one, a vote-based TOPSIS algorithm is proposed, which can efficiently reduce the subjectivity and avoide disturbing.For non-specific one, a directional distance based DEA algorithm is studied for optimizing all factors with linear programming. It transfers the issue into linear programming which greatly decrease the computational complexity with a little imprecision compared with globle searching and reinforce learning.5. Combined data and application layer, firstly, an integrated user trust scheme containing vitual identity, time, location, service classification and access history is established, which can solve the problem of trust’s retroaction caused by terminal hijacking. Then, in the sensor layer, smart antenna is utilized to solve the problem of sensor’s signal leakage. And a MDSM is introduced to conquer the shortcoming of normal fast-moving nodes localization method with DoA estimation. It has much better performance than MUSIC in scattering signal model generated by fast moving effect. |
PDF Full Text Request