SUM framework for optimizing shared on-demand fleet management

SUM framework for optimizing shared on-demand fleet management

The SUM project partners are developing a holistic framework to assist municipal governments and system operators in configuring new ride-pooling services. This comprehensive framework comprises three key components: a demand forecasting model, a purpose-built simulator for ride-pooling services (FleetPy), and an external optimization layer to identify optimal configurations.

Configuring an on-demand service involves numerous variables, such as the number of vehicles to operate, pickup and drop-off policies, and whether trip requests are processed in real-time or on a fixed schedule. Given the countless possible configurations, it is impractical to simulate all potential scenarios to find the best one for each case. The SUM project partners are investigating methods to maximize the efficiency of every simulation run.

SUM thumbnails website (4)

First, the project plans to feed simulation results back into the demand forecasting model. If the current service configuration performs poorly, fewer travelers will choose the service in subsequent simulations. Conversely, if it performs well, more travelers will opt for the service. Second, the team is exploring algorithms that can utilize previous simulation results to intelligently decide which configurations to try next. Once the basic framework is operational, the team will investigate ways to estimate system performance more quickly using established tools such as ExMAS.

The project includes case studies in the Living Labs of Jerusalem, Krakow, Munich, and Coimbra. In the Jerusalem case study, significant progress has been made on the demand forecasting model, which uses empirical travel data to generate plausible random trip requests with specific times, origins, and destinations. A travel preference survey conducted in the city will enhance the accuracy of this model and provide an evidence-based initial assignment of travelers to various modes (walking, private car, SUM on-demand service, public transit, or a combination of modes).

Lessons learned from the Jerusalem case study will inform the remaining case studies. In Krakow, plans are underway to launch van-pooling services integrated with tram lines later this year. The proposed methods allow for evaluating the suitability of different city areas for participation in the service. Out of eleven initial candidates, the local team is now focusing its analysis on two areas.