Protocol Design, Testing and Diagnosis towards Dependable Wireless Sensor Networks.

摘要

This thesis investigates the protocol design, testing, and diagnosis of Wireless Sensor Networks (WSNs) to achieve dependable WSNs.;In the aspect of protocol design, we focus on the MAC (Medium Access Control) layer protocol design. First, we propose an efficient MAC protocol RAS (routing and application based scheduling protocol) for underwater acoustic sensor networks (UWASNs), a type of WSNs that are deployed in the water. Utilizing the medium propagation difference between UWASNs and terrestrial wireless sensor networks (TWSNs), RAS performs parallel transmissions which would definitely result in collisions in TWSNs. It schedules the transmissions with different priorities by allocating longer time to heavier-traffic sensor nodes. The priority mechanism also benefits the fairness performance. Second, we tackle the unreliability problem caused by the packet loss in UWASNs. Based on the previously designed RAS, we propose a reliable RAS called RRAS that obtains a tradeoff between the reliability and the efficiency. RRAS applies an ACK and retransmission mechanism that is different from the traditional one, so that it can maintain a comparable throughput while improving reliability.;Third, in the area of protocol testing, we design RealProct (reliable Protocol conformance testing with Real sensor nodes), a novel and reliable framework for performing protocol conformance testing in WSNs, i.e., testing the protocol implementations against their specifications. With real sensor nodes, RealProct can ensure that the testing scenarios are as close to the real deployment as possible. To save the hardware cost and control efforts required by testing with large-scale real deployments, RealProct virtualizes a network with any topology and generates non-deterministic events using only a small number of sensor nodes. In addition, test execution and verdict are optimized to minimize the number of test case runs, while guaranteeing satisfactory false positive and false negative rates.;Finally, we propose MDiag, a Mobility-assisted Diagnosis approach that employs smartphones to patrol the WSNs and diagnose failures. Diagnosing with a smartphone which is not a component of WSNs does not intrude the execution of the WSNs as most of the existing diagnosis methods. Moreover, patrolling the smartphone in the WSNs to investigate failures is more efficient than deploying another diagnosis network. During the patrol, packets exchanged in the WSNs are collected and then analyzed by our implemented packet decoder. Statistical rules are also designed to guide the detection of abnormal cases. Aiming at improving the patrol efficiency, a patrol approach MSEP (maximum snooping efficiency patrol) is proposed. We compare MSEP with a naive method, the greedy method, and a baseline method, and demonstrate that MSEP is better in increasing the detection rate and reducing the patrol time than other methods.;We perform extensive evaluations to verify the proposed techniques and algorithms, and the results confirm their advantages in achieving dependable WSNs.