Abstract

Urbanization increases issues such as traffic congestion, lack of parking spots, and underutilized vehicles. In recent years, mobility-on-demand (MOD) concept has been proposed to effectively mitigate these issues. However, a common issue with MOD research is the lack of precise traffic data for conducting transportation-related studies and improving the proficiency of MOD systems. This is mainly because of data privacy concerns, GPS device limitations or errors, and expensive infrastructures for collecting real-time traffic data. Given the constraints, traffic simulations could be a reasonable solution for simulating the dynamic MOD activities such as distributing vehicles in the cities of interest and mimicking their movement behaviours. Despite the features that existing traffic simulators provide, they are not designed to support MOD use cases explicitly. For instance, background traffic generated by these simulators mostly follows random algorithms and the traffic flow is not based on real traffic patterns of the region. Another issue could be the lack of integration APIs to accept user inputs while the simulation is running to adapt the behaviour of the simulation. In this thesis, a synthetic MOD data generation framework is proposed. This framework takes a map region, real traffic data, and service vehicles trip plan as input. Using the ARIMA machine learning algorithm, we could predict demand and generate background traffic, followed by simulating the service vehicles in the region. The proposed framework generates synthetic traffic based on real traffic patterns and then simulates the service vehicles' movements on the map. While the simulation is running, the framework monitors the vehicles and collects real-time trajectory data. This framework leverages the features of SUMO as a microscopic simulation engine. In addition, established HTTP APIs enable third-party integration and allow users to control vehicles and trips on the map before and during the simulation execution. The offered simulation features include and are not limited to, the importation of a trip plan for numerous vehicles and the update of vehicle destinations. In addition to integration APIs, the proposed framework provides a graphical user interface to facilitate simulation setup and execution. The provided user interface enables users to explore a map, specify a region on the map, and then choose it as a simulation boundary. Throughout the simulation, the software core captures and stores real-time data on vehicle movement in a database that might be utilized for mobility-on-demand research. This simulation framework returns comprehensive service vehicle trajectories, departure time, destination time, travel duration, route length, and service vehicle status. The proposed software is open-source and publicly available, and its capabilities could be improved for future study.