The Research On Lifetime Optimal Routing Protocols In Wireless Sensor Networks

A Wireless Sensor Network (WSN) is a self-organizing network comprising a large number of wirelessly connected sensor nodes. Sensor nodes collaboratively transmit the collected data to the sink node. For sensor nodes, they are usually powered by limited batteries and are usually deployed in a harsh environment. Thus, it is difficult to replace batteries once the sensor nodes have been deployed. The feature of constrained energy brings a great challenge to WSNs, and how to prolong network lifetime becomes the primary objective of all kinds of related technologies. In a sensor node, the wireless communication module is the main energy consumer. The design of routing protocol affects the network lifetime of WSNs significantly. Therefore, it is of great importance to research on the lifetime optimal routing protocol for WSNs.This dissertation focuses on the lifetime optimal routing protocols. It defines the network lifetime of WSNs from three aspects:the time until the first node is drained of its energy, the time until the first living node has no path to the sink, and the energy efficiency of the whole network. These three aspects are used to evaluate the performance of routing protocols in this dissertation. To achieve the optimization of network lifetime in all defined aspects, this dissertation utilizes intelligent algorithms to address the routing issue. In addition, for different application scenarios, a WSN may consist of a few nodes or thousands of nodes. For this reason, this dissertation designs flat, hierarchical and multi-sink lifetime optimal routing protocols. The main contributions of this dissertation are listed as follows:(1) It proposes a reinforcement-learning-based lifetime optimal flat routing protocol: RLLO.For a small-scale WSN, it is appropriate to use the flat routing protocol. The flat routing protocol is simple and robust, but poorly scalable. Therefore, it is more suitable for small-scale network scenarios. The proposed protocol RLLO makes uses of the superiority of reinforcement learning to optimize the routing selection, and during this process, it takes such factors as link distance, residual energy, and hop count into account. It strives to reduce and balance the energy consumption, and ensure the packet delivery.The performance of RLLO has been verified by the NS2simulation tool. Among all the flat routing protocols, the EAR protocol has a certain advantage in the optimization of network lifetime, and I-EAR makes further improvements over EAR. Thus, this dissertation compares RLLO with these two protocols. Compared with EAR, RLLO yields48%and45%longer lifetime in terms of the time until the first node is dead and the first node is isolated, and gains119%increase in the overall energy efficiency. Moreover, in these three aspects, RLLO also achieves an improvement of31%,32%and46%over I-EAR.(2) It proposes an ant-colony-based lifetime optimal hierarchical routing protocol: ACLO-PEG.The hierarchical routing protocol is scalable, and suitable for relatively large-scale WSNs. ACLO-PEG is a hierarchical chain protocol. At first, it divides the network into multiple regions. Inside each region, it constructs a chain covering all the nodes. During each round of communication, it selects a head node for each chain, and makes all the head nodes forming a master chain. The mechanism of multiple chains solves these problems of low reliability and low latency which exists in the protocol of single chain. Accordingly, it can increase the amount of packet delivery. When the network topology changes, the ACLO-PEG protocol just needs to partially reorganize the chains, which greatly reduces the cost of routing mamtainance.In order to solve the local optimal problem in the PEGASIS protocol, this dissertation puts forward the ACCH algorithm to construct the chain. ACCH takes advantage of ant colony optimization (ACO) to build a chain with minimum total energy consumption. Besides, relative to the ACO model, ACCH mends such aspects as the probability distribution and the pheromone updating. Besides, ACLO-PEG considers residual energy and distance to the sink to select the head of the master chain. Such a head node needs to directly communicate with the sink node. But all the other nodes only communicate with their neighbor nodes on the chain. Therefore, the ACLO-PEG protocol can decrease the total energy consumption to some extent, meanwhile balancing the energy consumption.Among the hierarchical routing protocols, PEGASIS is a classical chain protocol. The chain architecture of PEGASIS has inherent advantages. And H-PEGASIS is a hierarchical PEGASIS. To verify the performance of ACLO-PEG, this dissertation compares ACLO-PEG with these two protocols. The results show that, compared with PEGASIS and H-PEGASIS, ACLO-PEG gains an improvement of93%and35%in terms of the time until the first node is dead,80%and37%in terms of the time until the first node is isolated,28%and20%in terms of the overall energy efficiency.(3) It proposes a lifetime optimal routing protocol with multiple sinks:LOMS.For a large-scale WSN, the architecture of multiple sinks has advantage of network stability and network robustness. Then, a multi-sink routing protocol, LOMS, is proposed in this dissertation. LOMS distributes network traffic to multiple sinks to balance the energy consumption. Motivated by the theory of Minimum Cost Maximum Flow in operational research, LOMS utilizes the proposed algorithm ACSP to find the optimal path between sensor nodes and multiple sinks. ACSP is an ant-colony-based algorithm to search for the shortest path.This dissertation simulates and compares LOMS with MRMS. MRMS is one of the most representative lifetime optimal multi-sink routing protocols. Over a series of experiments this dissertation comes to the conclusion that LOMS shows superior performance when there are four sinks in WSNs. In this scenario, LOMS yields59%and48%longer lifetime over MRMS in terms of the time until the first node is dead and the first node is isolated. In addtion, in terms of the overall energy efficiency, LOMS also gains an improvement of62%.For these three proposed routing protocols RLLO, ACLO-PEG and LOMS, this dissertation finally tests and compares their performance under a variety of network scenarios. The results show that the flat routing protocol RLLO has a poor scalability. In a small-scale network scenario, RLLO shows superior performance. Howevr, with the expansion of network scale, the performance of RLLO declines fiercely. When the number of nodes in the network is more than400, ACLO-PEG has the best performance in the first aspect of network lifetime. However, when the number of nodes reaches600, LOMS shows distinct advantages. In conclusion, the flat routing protocol RLLO is suitable for the small-scale WSNs, the hierarchical routing protocol ACLO-PEG is applicable to such applicaitons which require a relatively large network scale and a full coverage of sensor nodes, and the multi-sink routing protocol LOMS suits the large-scale WSNs.