Introduction: This project is a collaboration between electrical and computer engineering (Dr. Pompili), and civil and environmental engineering (Dr. Hill) on sensing and modeling extreme weather events. Such events have profound effects on the sustainability of urban centers. At the same time, human activities are increasing the variability of the climate and increasing the frequency of these events, driving the need for dynamic decision-making tools.
The overall objective of this research is to address urban sustainability through the development of modeling methods suitable for forecasting environmental phenomena in a changing world, and through the development of technology that can enable autonomous infrastructure to adapt to rapidly evolving environmental conditions. This research will focus on rainfall estimation and measure success by the ability to provide accurate rainfall estimates at resolutions higher than the minimum threshold suggested by the literature.
Motivation: Currently all of the rainfall data is acquired either by radar or by satellite. These techniques have been successfully used for many years. However, both radar and satellite predict rainfall only over large areas, and quite often the actual precipitation varies largely on a higher spatial resolution. One way to solve this problem and get accurate data about precipitation is to place rain sensors directly in the area being considered. Those rain sensors can provide a very accurate online reading and help in the early detection of flooding. Unfortunately, with quality comes a high price tag. The proposed solution is to take advantage of ubiquitous rain sensors in vehicles. Since 2006, many vehicles began integrating simple precipitation sensors to automatically turn on the wipers. Those sensors are cheap and ubiquitous, yet they are not very accurate. The idea however is to take advantage of the multitude of those sensors, being deployed on more and more vehicles every year. The high density of the sensors is expected to compensate for lack of precision, and still deliver useful data.
Design: A network of portable and self-sufficiently powered sensors capable of transmitting data for a long time in a desired area was constructed. Each sensor is able to collect and transmit data to its parent router, which is then sent up the hierarchy on the network for further processing. Sensing units are equipped with solar cells regenerating power supply. Web-based analysis of data accessible to the public was developed. Both real time and previously recorded data from any active sensor unit can be displayed with desired time frame and measurements. A distributed adaptive sampling protocol, SILENCE, was implemented to reduce transmission overhead and to prevent sending repetitive data from multiple sensors on the central network by leveraging data redundancy due to spatial correlation and similarity.
Conclusion: This research focused on creating a real-time system for rainfall adaptive sensing using ubiquitous sensors, which will permit to pursue the ambitious overall objective of enabling predictive control of urban infrastructure during extreme weather events.