| Internet of Things (IoT) is an organic interconnected syntheses of the agent, the intelligent control and communication relationship between the agents. It is changing people’s production and life and will cause an important information revolution. IoT combines many fields of science and technology, which include wireless sensor networks, embedded system, intelligent control, data processing and merging. For the large synthesis, how to evaluate the performance in the design step and feedback to guide the design, become a research focus.This dissertation presents modling analysis methods based on queueing networks and performance optimization strategy for IoT. The queueing network model with blocking is established for large scale sensor nodes in IoT. The path delays, configurations of the optimal buffer size, design of sink node based on embedded multi-core SoC and improving the node positioning accuracy are in-depth study. Based on the above research, the graphical tool software of queueing network modeling and simulation is developed. These works provide the important reference to the pre-design and design evaluation for the building IoT. The main contributions of the dissertation are as follows:1. The delay analysis modeling method based on open queueing network for large scale sensor nodes is proposed. By the analysis method of nodes’ end-to-end delay and average delay of whole queueing network, the multi-path selection mechanism is determined. An iterative approximation algorithm is proposed for the qualitative analysis of the packet arrival rate of sensor nodes. In order to reduce the complexity of multi-path delay calculation, the pre-selection algorithm based on path search tree is designed. The redundant paths are removed, the number of pre-selection path is much less than the actual number of paths. The optimal path and alternate path are obtained by the end-to-end delay analysis method. The method provides an effective solution for selecting the data transmission path and improving the data transmission efficiency in the large-scale wireless sensor network.2. The best configuration method of hardware packet buffer for the different types of sensor nodes is proposed. The packet queueing network model is established for large-scale sensor nodes using queueing networks with blocking. In order to analyze the blocking situation, the finite holding nodes are added to an open queueing network of the wireless sensor networks. Equivalent queueing network model is obtained. According to the usage of holding nodes, packet buffer size is determined. The consistency of model calculations results and statistical experiments measuring results for wireless sensor networks of the value are verified by comparing the calculated and experimental measurement results. The method provides a theoretical modeling basis for hardware buffer configuration and optimization of the designing high cost-effective in large-scale sensor networks.3. According to the characteristics of the distributed data in the IoT, the modeling and performance analysis method for sink node based on embedded multi-core SoC has been proposed using queueing networks with different priorities. Each executing core is assigned different buffers with priorities. According to the blocking circumstances of each executing core, scheduling is realized. Embedded multi-core SoC performance is greatly improved. The evaluation algorithm of queueing network model is fulfilled for embedded multi-core SoC. When the system is able to get the best system performance, the optimal values of the required hardware buffer capacity are set. By comparing task arrival rates before and after application of adaptive scheduling algorithm, adaptive scheduling algorithm proportion task allocation of the executing cores is proved.4. Traditional RSSI-based positioning method is improved and the positioning accuracy is improved.The N-time trilateral centroid weighted localization algorithm (NTCWLA) is proposed, which can reduce the error considerably. Considering the instability of RSSI, the weighted average of many RSSIs are used for current RSSI. In order to improve the accuracy, a number of reliable beacon nodes are selected to increase the localization times. The mobile node is real time located N times using NTCWLA. The results show that the proposed algorithm performs better than the trilateral centroid algorithm. This method provides an important technical support related to the positioning of the IoT and smart mobile body monitoring.5. The simulation tool software for performance evaluation of Things is developed. The simulation software based on queueing network modeling is designed. A typical simulation performance analysis example for IoT node topology is tested. Actual hardware testing environment of IoT is build. According to comparison of the simulation data and actual test data based on hardware, the results of the designed simulation modeling software are basically consistent with statistical results of actual hardware applications for IoT. These constructions provide important references for performance evaluation of IoT... |
PDF Full Text Request