Energy-saving cyber-physical systems in smart cities

The ubiquity of mobile devices and maturing inter-connecting technologies have facilitated numerous smart services pervading every aspect of our urban life. Combining Internet-of- Things (IoT) technologies with a suite of machine learning and analytic tools, we are witnessing the era of intelligent cyber-physical systems (CPS) that rapidly transform the way of living. One of the primary goals of CPS is to make our cities more efficient and sustainable, by utilizing multi-modal information, optimizing resource planning and improving the interactions with the physical world. This thesis focuses on intelligent transportation system, which is an important aspect of smart city services that expects significant improvements. We make several contributions towards the goal of fuel-saving in modern transportation, exemplified by three major systems. The first system called GreenRoute offers fuel-saving routes instead of the fastest routes. It works by exploiting road features existing in public map data sources to train fuel consumption models that enjoy extendability to many cities. GreenRoute solves vehicle heterogeneity problem using a normalization process, which lowers maintenance cost and software complexity. The second system named GreenDrive comes into effect after the optimal route is chosen. GreenDrive is a type of Green Light Optimized Speed Advisory (GLOSA) service, which offers optimal driving speed based on real-time traffic signal states. It strives to reduce the occurrence of stop-and-go movements at signalized intersections, by learning signal phase and timing from vehicular sensing data and predicting future signal schedules. The third system is CoDrive, which is a coordination service aiming at a global fuel-saving optimality for the platoon as a whole. CoDrive works on top of GreenDrive, with a cost estimation module calculating the combined fuel consumption of involved vehicles after sharing their individual optimality. Finally, we put these systems together and create realistic simulation scenarios on real maps of four representative regions of Chicago. Real traffic datasets are used to generate and validate traffic flows, and representative time periods are chosen to estimate daily and weekly savings. Using population data, finally we quantitatively derive how much fuel can be saved in the city of Chicago per year.LimitedAuthor requested closed access (OA after 2yrs) in Vireo ETD syste