The convergence of computation, communication and sensing has led to the emergence of Wireless Sensor Networks (WSNs), which allow distributed monitoring of physical phenomena over extended areas. In this thesis, we focus on a dual flood and traffic flow WSN applicable to urban environments. This fixed sensing system is based on the combination of ultrasonic range-finding with remote temperature sensing, and can sense both phenomena with a high degree of accuracy. This enables the monitoring of urban areas to lessen the impact of catastrophic flood events, by monitoring flood parameters and traffic flow to enable public evacuation and early warning, allocate the resources efficiently or control the traffic to make cities more productive and smarter. We present an implementation of the device, and illustrate its performance in water level estimation and rain detection using a novel combination of L1 regularized reconstruction and machine learning algorithms on a 6-month dataset involving four different sensors. Our results show that water level can be estimated with an uncertainty of 1 cm using a combination of thermal sensing and ultrasonic distance measurements. The demonstration of the performance included the detection of an actual flash flood event using two sensors located in Umm Al Qura University (Mecca). Finally, we show that Lagrangian (mobile) sensors can be used to inexpensively increase the performance of the system with respect to traffic sensing.
These sensors are based on Inertial Measurement Units (IMUs), which have never been investigated in the context of traffic ow monitoring before. We investigate the divergence of the speed estimation process, the lack of the calibration parameters of the system, and the problem of reconstructing vehicle trajectories evolving in a given transportation network. To address these problems, we propose an automatic calibration algorithm applicable to IMU-equipped ground vehicles, and an L1 regularized least squares formulation for vehicle speed estimation. Results show that this system can be used to generate accurate traffic monitoring data, and significantly outperforms GPS sensors (traditionally used as traffic flow sensors) in terms of cost, accuracy and reliability.
|Date of Award||Oct 2016|
|Original language||English (US)|
- Computer, Electrical and Mathematical Science and Engineering
|Supervisor||Christian Claudel (Supervisor)|
- flood detection
- Wireless Sensor Networks
- mobile sensors
- Machine Learning