Publication: Extending wireless sensor network lifetime using battery state of charge estimation and media access control layer improvements
Loading...
Date
2023-01-01
Authors
Omer Ali
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Wireless Sensor Networks (WSNs) are comprised of many low-power, constrained sensor nodes that are typically deployed in close proximity to perform various sensing and data acquisition tasks. These devices are powered by batteries in most cases, with limited on-board energy (capacity). Therefore, a sensor node's lifetime is typically defined by the end of its batteries. Furthermore, operational conditions (such as load) and environmental (such as temperature and humidity) conditions can have a direct impact on the available battery capacity. Typically, WSN development is usually performed in network simulators that can only estimate a small number of operational parameters, resulting in inaccurate real-world estimates. The inability to monitor the external effect on WSN node batteries is rarely reported resulting in major energy estimation inaccuracies, that further reflects on every layer’s operation. In addition, Clear Channel Assessment (CCA) based enhancements were previously suggested duty-cycle based MAC protocols as being computationally efficient, and simpler to implement; however, due to the fixed schedule, they limit scalability as the network grows. To address these issues, a novel cross-layer approach is presented that is computationally efficient, platform independent, and can be extended to existing techniques. To begin with, a thorough battery characterization was performed to assess various battery chemistries. In this regard, a novel five-step methodology was proposed suggesting the optimal battery operating conditions for WSN applications. Next, the battery discharge profiles were analysed for feature selection and mapping to machine learning algorithms for battery state of charge (SOC) estimation. In this regard, a lightweight Gaussian Process Regression based Radial Bias Filter (GPR-RBF) technique was proposed that can achieve SOC estimation accuracy of over 98 percent for all battery types and operational conditions. Next, a comprehensive testbench was proposed to evaluate the accuracy of the proposed scheme with existing techniques, for accuracy and scalability. Finally, adaptive clear channel assessment (A-CCA) was proposed and evaluated for performance gains. It was concluded that with A-CCA, the transmission range, packet delivery rate (PDR), and packet reception rate (PRR) were increased without incurring additional duty-cycle cost. Sensor node lifetime was reported to have increased by 12% because of the combined cross-layer optimizations.