Research On Performance Modeling And Optimization Design In Wireless Sensor Networks

A wireless sensor network (WSN) consists of a number of integrated sensor nodes which are monitoring, sensing and collecting information about the environment or the monitored objects in a cooperative manner. The information is transmitted to the user terminal via wireless multi-hop relay communications, which can be regarded as a form of"pervasive computing". These unique features enable a wide range of WSN applications, such as military affairs, national security, habitat monitoring, traffic management, antiterrorist battle, disaster prediction, and health care, etc, which has attracted much attention in the academia and industries research areas.Recently, as more and more WSN applications are getting developed, there is even increasing demand in the performance of network protocols (including the MAC protocol, routing protocol, etc) in WSN. This requires us to model and analyze the network protocols, so that they can be effectively designed and optimized. To achieve this goal, we need to find theoretical tools and develop innovative methods to accurately analyze the network protocols and its impact on network performance. This thesis mainly focuses on the performance modeling and analysis methods in WSNs, and validates the accuracy of network models. At the same time, it proposes some network protocols optimization and design algorithms, and analyzes their performance in detail, then validates them through large numbers of computer simulations. The main contents and contributions of the thesis are listed as follows:(1) For a contention-based WSN, each node's behavior is modeled as an M/M/1/K queue, and a general model with low complexity is proposed for the contention-based network. With this model, the relationship between system parameters (e.g. duty cycle and average active time) and network performance (e.g. the average power consumption, average packet delay, average packet loss rate and network throughput) is obtained and given a qualitative theoretical analysis. The models provide us an elementary way to theoretically analyze the impact of network parameters on the network performance in WSN.(2) For commonly-used contention-based WSNs with synchronous wakeup patterns, the modeling methods are extended to propose a low-complexity and fairly accurate model that takes the synchronicity into account. It can help us to study the impact of system parameters on network performance. This work effectively improves the extension of our modeling method. (3) For SMAC, a classical MAC protocol used in WSNs, a network performance modeling framework is proposed under unsaturated conditions. It accurately models the key features of SMAC such as the periodical active/sleep mechanism, fixed contention window size, random backoff, hidden terminal problem and virtual clusters to set up the network model. An iterative algorithm is proposed to compute the channel contention probability and service time distribution for each node. Using these results, network performance such as throughput, delay and energy consumption can be predicted. The modeling framework thus provides us a good way to study the dependency of network performance on the network parameters such as duty cycle and data traffic, and optimize the system parameters to obtain better network performance.(4) By introducing the idea of cross-layer optimization in the network design, the optimization of transmission power allocation in PHY, packet retransmission in the MAC layer and routing path selection in the network layer are considered jointly. With the objective being the maximization of network lifetime under a packet timeliness constraint, a cross-layer strategy is proposed, which consists of a centralized transmission power allocation algorithm (Centralized TPA) which minimizes the path energy under the end-to-end packet timeliness constraint, and an optimal routing path selection algorithm (ORPS), which maximizes network lifetime by optimizing the routing path fractions.(5) The Centralized TPA and ORPS in (4) have fairly high algorithm complexity and communication overhead. To alleviate these problems, low-complexity distributed algorithms including the Lagrange-dual-based TPA and cost-based RPS algorithms are proposed. These algorithms have fast convergence speed and can converge close to the optimal performance, and they are particularly suitable for the design and implemention of practical large-scale WSNs.