,
Dynamic network-level models directly addressing ride-sourcing services can be useful for the development of efficient traffic management strategies both for city and company operators. Recent developments presented models under equilibrium situations for several ride-sourcing service settings and dynamic models focusing on ride-hailing (solo rides), but no work addressed ridesplitting (option for shared rides) in dynamic settings. In this work, we sought to develop a dynamic aggregated network model capable of representing ride-sourcing services and private vehicles macroscopically in a multi-region urban network. To address this, we combined the use of Macroscopic Fundamental Diagram (MFD) with a detailed state-space and state-transition description embracing private vehicles and ride-sourcing vehicles in their several activities to formulate adequate mass conservation equations. Accumulation-based MFD dynamic models might experience additional errors due to the strong variations of trip lengths, e.g. when vehicles are cruising for passengers. We integrate the so-called M-model that utilizes one additional set of state variables, the total remaining distance to be traveled for a region. We show that the model can accurately forecast the vehicles’ conditions in short-term predictions (up to 30 minutes ahead of time). Later, a comparison with a benchmark model showed lower errors in the proposed model in all states. The development of such a model prepares the path towards the development of real-time feedback based management policies such as repositioning strategies for idle ride-sourcing vehicles and the development of regulations over ride-sourcing in congested areas.
2022
