Abstract: A Wireless Sensor Network (WSN) system consists of sensor nodes that can sense and transmit physical information of environments or objects being monitored. Infrastructure monitoring is an important class of WSN applications, where a WSN is used to continuously monitor and assess the health status of an infrastructure structure with only little human activity. An infrastructure monitoring WSN enables early detection of problems, further activating the precautionary mechanism to ensure public safety and environment protection. For example, a damaged segment in a railroad track could lead to serious human injures and property loss; the early detection of the damaged segment can prevent major accidents. Similarly, an early detection of leaks of hazardous material such as crude oil and natural gases can prevent accidents, such as pollution and the eruption of fires.
   Due to the limited resources of sensor nodes, WSN systems may face many challenges, such as energy consumption, communication quality and data processing efficiency. Data fusion provides an important technology to address these problems with great potential in achieving efficient information gathering and analysis, and energy utilization and communication reliability have been improved significantly. Reliability analysis, as an essential indicator of the system’s stability, plays a pivotal role in reducing the occurrence of failures, ensuring the system’s consistent operation, enhancing the system’s performance and reducing maintenance costs. Consequently, research on the reliability of low-energy WSN systems has gradually become a hot topic.
   Modern WSN systems often use a large number of sensor nodes to perform a set of information gathering and analysis computations. The sensor nodes are often heterogeneous due to factors such as different suppliers, model types and operating environments. In addition, these sensor nodes typically exhibit more than two performance levels or states corresponding to different computing powers and demands.
   Reliability modeling and analysis of large-scale WSN systems is difficult. State-space-based methods like Markov or semi-Markov chains are potentially applicable. However, they suffer from the state-space explosion problem when analyzing medium or large-scale systems, and are typically limited to integrable time-to-failure distributions. On the other hand, discrete event simulations can be used to handle arbitrary types of distributions, but they generally require a lot of computational time and can only offer approximate results.
   Recently, an efficient algorithm based on the Multi-valued Decision Diagram (MDD) has been proposed to analyze multi-state systems. MDDs are efficient graph-based data structures for symbolic representation and manipulation of multi-valued logical functions. Based on Shannon’s decomposition theorem, MDDs can represent multi-valued logical functions as a rooted and Directed Acyclic Graph (DAG) in the form that is both canonical and compact through two reduction rules, “merging isomorphic sub-trees” and “deletion of useless nodes”. MDDs have provided an efficient method for reliability analysis of diverse systems, such as multi-state systems, phased-mission systems and large-scale networks.
   In this work, we make new extensions of the MDD model for analyzing low-energy WSN systems with heterogeneous and multi-state components. Firstly, the system and reliability problem of low-energy WSN based on PEGASIS protocol are defined. Then a top-down approach to generate a MDD model based on data fusion is proposed, which uses the several truncation and merging operations to avoid the generation of a large number of unnecessary MDD nodes, resulting in a compact MDD model. Finally, the system reliability evaluation is carried out based on the constructed MDD model. Once the MDD model is constructed, the reliability of the system can be analyzed and evaluated based on different failure time distributions.
   The constructed MDD model can be used to efficiently calculate the reliability of a low-energy WSN and the proposed method is illustrated through a specific example, and the results show that the proposed MDD method can solve the reliability evaluation of low-energy wireless sensor network systems with different configurations and failure parameters of sensor nodes. Moreover, compared to traditional reliability evaluation methods, the proposed method significantly improves the efficiency of reliability calculation. The MDD method can effectively alleviate the combinatorial explosion problem of enumeration methods and perform efficient reliability analysis of low-power WSN systems.
   The reliability analysis is the foundation of low-energy WSN system optimization. The results of reliability analysis can provide reliable data support and result analysis for actual industrial projects. The results of reliability analysis can be used by the system designer to obtain the optimal design of the network topology and node deployment strategy to meet some of his needs in industry and make cost-effective and optimal decisions on the number of sensor nodes configured in the system and the resources allocation of the sensor nodes themselves. Consequently, our future work focuses on related optimization problems, such as the optimization of energy management and chain structure in sensor nodes, with the objective of further enhancing the availability and survivability of low-energy wireless sensor network systems.

Key words: low-energy wireless sensor network system, data fusion, reliability analysis, multi-value decision diagram

摘要： 低功耗无线传感器网络(Wireless Sensor Networks, WSN )在基础设施监测和环境监测等方面有着十分重要的应用,它对信号覆盖范围区域的特定活动对象进行信息特征的识别、采集和提取,并通过无线通信的方式进行自组织的传输。针对异构多状态的低功耗无线传感器网络系统可靠性分析问题,提出了一种基于多值决策图(Multi-value Decision Diagram,MDD)的高效可靠性分析方法。首先对以PEGASIS协议为代表的簇链结构WSN进行可靠性问题定义并采用MDD进行建模,然后根据感知数据传输和融合特征设计多维度的约简规则用以生成高效且紧凑的MDD模型,最后基于构造的MDD模型高效地进行系统可靠性评估。通过具体实例对提出的MDD方法进行验证,实验结果表明,通过采用多个维度约简规则的MDD方法能够有效缓解枚举类方法的组合爆炸问题,对低功耗WSN系统进行高效可靠性分析。

关键词: 低功耗WSN, 数据融合, 可靠性评估, 多值决策图