The capability of sensor nodes has been improved tremendously in the last decade. However, wireless sensor network (WSN) protocol performance still not considered reliable, predictable, or even repeatable. Although performance repeatability is a fundamental requirement in the network protocol development, in many cases, wireless protocols, which are extensively tested and well tuned in the WSN testbeds, misbehave in the actual target environments. In this dissertation, we study how to achieve performance repeatability of WSN protocols. There are a couple of crucial reasons why WSN protocol performance is not considered repeatable: unreliable hardware, software, and low power radio communication. Moreover, network protocol behaviors can themselves exhibit nontrivial variability, and this variability may only be inadequately understood in the testing phase. The multi-faceted difficulty with ensuring desired protocol behavior in the field coupled with the high cost of testing and tuning the performance in the field, motivates the scientific study of tools and techniques for reproducing network behavior across test and deployment environments. To attack this performance unrepeatability problem of WSN protocols, we approach solutions with analytical and data driven methods. With analytical method, we characterize the wireless protocol performance and behavior mathematically. We try to identify uncertainty factors of WSN environments such that WSN testbeds and deployments. In the next step, we analyze their impact on protocol performance and behavior. Towards achieving reproducible performance across networks of potentially different environments, we adopt the concept of realizing the same (or measurably close to the same) ``link usage spectrum'', defined as the probability distribution with which the network protocol selects links of different length from among all the available links in the network at hand. Based on the mathematical modeling of link usage spectrum, we derive a closed form equation of the expected performance and the variance of wireless protocols using PRRxD for routing metric. Equipped with mathematical modeling of link usage spectrum and the expected performance of wireless protocols, we provide methods to reproduce the comparable protocol performance across environments by matching link usage spectrums as close as possible (method 1) or by matching the expected performances as close as possible (method 2) of two different environments. These two methods work well with 1-dimensional chain topologies and 2-dimensional grid topology, but are not applicable to topologies except chain or grid. This problem can be solved with data driven method. With data driven method, we simulate wireless protocols with operational models with detailed link quality data collected with a WSN testbed resource specification profiling program, RS-Profiler. It is difficult to analytically model effects such as multi-path and component variability. However, models based on measurement data that captures these effects can improve accuracy substantially. We implement RS-Profiler that collects RF data of all links and all nodes, e.g. RSSI, SNR, PRR, noise floor, efficiently. We provide two performance prediction algorithms (operational models of routing protocols) that accurately predict the expected performance of protocols using the cumulative routing metric (e.g. ETX) and the 1-hop routing metric (e.g. $PRR× D$) based on the collected RF resource specification. These two algorithms can be applied to any topologies. We prove that performance repeatability within WSN testbed and also across WSN testbed with performance prediction algorithms and RF resource specification with extensive experiments over 18 different 2-dimensional grid networks. However these data driven methods incur scalability problems in data collection. We present a study on time complexity of RS-Profiling, which is $O(N)$ with the number of nodes N. Because the profiling time will be impractically long (e.g. 3 days for full RSSI, noise floor, and PRR for all 16 channels, for all 8 different transmission power levels, for all links of all nodes), we present studies of three methods to relieve this RS-Profiling time scalability problem.
|Commitee:||Ertin, Emre, Koksal, Can, Srinivasan, Kannan|
|School:||The Ohio State University|
|Department:||Computer Science and Engineering|
|School Location:||United States -- Ohio|
|Source:||DAI-B 78/11(E), Dissertation Abstracts International|
|Keywords:||Performance calibration, Testbed, Wireless sensor network|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be