Mobile systems for adaptive road safety and time-relevant modal shift

Mobile networks have brought the opportunity of using smartphone data to investigate the challenges of road safety and travel mode choice for sustainable urban mobility. This thesis adopts the mobile cloud computing (MCC) approach and uses smartphone data to address these two challenges. Road safety influences the mode choice between private cars and low-carbon transportation that usually involves walking and cycling. In this regard, pedestrian collision avoidance is crucial for both conventional and autonomous vehicles. As dedicated 802.11p-based device-to-device communication is not readily available on smartphones, collision warning systems could use a designated mobile app for indirect vehicle-to-pedestrian (V2P) communication and centrally perform the collision prediction on cloud servers. Vehicular safety applications require high-frequency beaconing to achieve adequate locational and temporal precisions. In these situations, the power consumption of beaconing can create a bottleneck due to the limited capacity of smartphone batteries. In addition, cellular network capacity and cloud computation resources could be limiting factors that require their own city-scale assessments. To address these issues, this thesis investigates the practicality of smartphone-based V2P collision warning systems by means of model and prototype evaluations. We need to analyze extent and variations of battery consumption under different traffic conditions, assuming a situation-adaptive rate control is used for communication. We also need to examine the network and computation load of the system and compare them against the capacity and cost of available infrastructures. Regarding mode choices, a modal shift from cars to lower-carbon alternatives results in changes in travel-times, emissions, and physically-active distances. An increased travel-time can be a barrier against the realization of modal shifts, due to limited daily time-budgets of people. Despite significant development in data collection technologies, there is a limited focus on the decision-support perspective and developing useful computational frameworks for a time-relevant modal shift analysis. This thesis aims at developing a prototype of an MCC-based framework that uses smartphone data to show the extent of potential changes with modal shifts while taking into account travel-time changes. Such a framework would involve extracting observed door-to-door trips from smartphone data, computing realistic low-carbon alternatives to compare with those observed trips, and finally estimating the potential changes. Moreover, due to the nature of urban travel patterns, any proposed method should be evaluated in realistic scenarios and consider spatiotemporal variations. This thesis seeks to acquire realistic trace data of target urban regions and use it to evaluate the proposed frameworks. While the numerical outcomes could differ between cities, such experimental evaluations should demonstrate the overall practicality and usefulness of the methods.