Abstract
Monitoring and control systems have played a central role in industry and everyday life, often in a non-intrusive manner. Yet, they will become more ubiquitous, autonomous and distributed, with the rapid development of the envisioned technologies of "smart homes", "smart cities" and "industry 4.0". All these new technologies are built upon the enabling technology of WSN. Their success depends to a large extent on the communication capability of WSNs. Fast reaction and feedback is a common characteristic of these technologies, therefore, how to achieve low-latency, high-reliability and flexibility in WSN communication is a key challenge, and a decisive success factor.
WSN refers to a network that connects a number of low-cost, low-power sensor nodes which have sensing and/or actuating capabilities, and can communicate with each other over short distance via a low-power radio. The predominant advantages that WSN offers are 1) distribution and fault tolerance in communication and sensing/actuation by leveraging large number of sensor nodes, 2) cost reduction by removing the cables of communication and power supply, and 3) flexibility in the deployment of tiny cableless sensor nodes. However, one main drawback of the wireless technology, in contrast to the mature wired counterpart, is the much weaker communication capability --- the combined result of stronger interference in the wireless channel, the weak signal strength of low-power radios and the complexity in the scheduling of multi-hop wireless communication.
The main goal of the thesis is to facilitate the transition from wired technology to wireless technology for industrial automation. Specifically, I provide solutions for improving and guaranteeing QoS in WSN communication.
I tackle the problem for two scenarios where the network topology is either known or not. When the network topology is known, I adopt an approach of reservation-based scheduling, i.e., through centralized scheduling of communication opportunities, in order to optimize various communication metrics. In the thesis, I propose a very efficient multi-channel scheduling algorithm that gives nearly optimal latency performance (within 1.22% of the optimum) for the tree-based convergecast, which is by far the predominant communication pattern, especially for monitoring applications. I also propose very efficient multi-channel scheduling algorithms that offer high schedulability and low overhead for multi-flow periodic real-time communication on an arbitrary network topology with multiple gateways. Such a communication pattern is typical of a multi-loop control system.
On the other hand, if the network topology is unknown or changes very dynamically, I optimize the QoS in communication by exploiting concurrent transmission on the physical layer, which is routing-free by nature. First, I proposes a simple model for concurrent transmissions in WSN which accurately predicts the success or failure in the packet reception. Then I design the Sparkle protocol for highly reliable, low latency and energy efficient multi-flow periodic communication.
Finally, it presents the Ripple protocol for high throughput, reliable and energy efficient network flooding using pipeline transmissions and forward error correction, which significantly improves the state-of-the-art.
Although the thesis assumes WSN as the communication technology and industrial automation as the application scenario, it is by no means restricted to these settings since the proposed solutions can be applied to other wireless networks and other scenarios with similar communication patterns and QoS concerns.