| dc.description.abstract | Reliable, real-time traffic surveillance is an integral and crucial function of the 21st century intelligent transportation systems (ITS) network. This technology facilitates instantaneous decision-making, improves roadway efficiency, and maximizes existing transportation infrastructure capacity, making transportation systems safe, efficient, and more reliable. Given the rapidly approaching era of smart cities, the work detailed in this dissertation is timely in that it reports on the design, development, and implementation of a novel, fully-autonomous, self-powered intelligent wireless sensor for real-time traffic surveillance. Multi-disciplinary, innovative integration of state-of-the-art, ultra-low-power embedded systems, smart physical sensors, and the wireless sensor network—powered by intelligent algorithms—are the basis of the developed Intelligent Vehicle Counting and Classification Sensor (iVCCS) platform. The sensor combines an energy-harvesting subsystem to extract energy from multiple sources and enable sensor node self-powering aimed at potentially indefinite life. A wireless power receiver was also integrated to remotely charge the sensor’s primary battery. Reliable and computationally efficient intelligent algorithms for vehicle detection, speed and length estimation, vehicle classification, vehicle re-identification, travel-time estimation, time-synchronization, and drift compensation were fully developed, integrated, and evaluated. Several length-based vehicle classification schemes particular to the state of Oklahoma were developed, implemented, and evaluated using machine learning algorithms and probabilistic modeling of vehicle magnetic length. A feature extraction employing different techniques was developed to determine suitable and efficient features for magnetic signature-based vehicle re-identification. Additionally, two vehicle re-identification models based on matching vehicle magnetic signature from a single magnetometer were developed. Comprehensive system evaluation and extensive data analyses were performed to fine-tune and validate the sensor, ensuring reliable and robust operation. Several field studies were conducted under various scenarios and traffic conditions on a number of highways and urban roads and resulted in 99.98% detection accuracy, 97.4782% speed estimation accuracy, and 97.6951% classification rate when binning vehicles into four groups based on their magnetic length. Threshold-based, re-identification results revealed 65.25%~100% identification rate for a window of 25~500 vehicles. Voting-based, re-identification evaluation resulted in 90~100% identification rate for a window of 25~500 vehicles. The developed platform is portable and cost-effective. A single sensor node costs only $30 and can be installed for short-term use (e.g., work zone safety, traffic flow studies, roadway and bridge design, traffic management in atypical situations), as well as long-term use (e.g., collision avoidance at intersections, traffic monitoring) on highways, roadways, or roadside surfaces. The power consumption assessment showed that the sensor is operational for several years. The iVCCS platform is expected to significantly supplement other data collection methods used for traffic monitoring throughout the United States. The technology is poised to play a vital role in tomorrow’s smart cities. | en_US |
