Over the last decade, on-demand mobility services have drastically transformed the urban transportation ecosystem. Driven by the rise of new technologies, especially mobile networks and terminals, the number of Transportation Network Companies (TNCs) on the market has multiplied, competing with traditional taxi companies and providing travelers with a vast range of services. These services can meet an increasingly dynamic and irregular mobility demand, unsatisfied by public transportation or personal car constraints. On the one hand, ride-sourcing companies offer more flexible services than public transit, on-spot and on-demand pickup, and no connections. On the other hand, it can be less costly than private car ownership and provides satisfying solutions to parking issues.
These individual benefits extend to a collective scale, at which the limitation of car ownership can relieve the pressure on land and the need for parking spaces. By encouraging pooled trips through ride-splitting, these services can reduce road traffic and its externalities: participation in urban congestion, air emissions, or noise pollution. Optimally articulated with public transportation networks, these services can enhance accessibility to urban services from suburban and rural areas. Yet, efficient management of this type of service requires successfully handling several operational issues: demand prediction, ride-pricing, passenger-vehicle matching, vehicle routing, and fleet rebalancing. The newly published DIT4TRAM report addresses this latter issue.
Urban mobility dynamics and asymmetry are significant issues in managing on-demand mobility services. Mobility-on-demand fleets are affected by spatial and temporal patterns of demand. Without fleet rebalancing strategies, vehicles accumulate in the most attractive areas of the network, to the detriment of the high-demand regions which suffer from vehicle depletion. The satisfaction of new ride requests in those areas may require vehicles to travel long distances to pick passengers up, implying substantial waiting times for the service users. These waiting times affect the attractiveness and performance of the service and may lead passengers to abandon their ride.
On the contrary, anticipating future demand and relocating the fleet accordingly ensures high service levels to users and minimizes abandonment rates. Thus, two questions arise: how to predict ride-sourcing demand and how to reorganize the fleets accordingly?
DIT4TRAM report research topics
The report focuses on the latter issue. Fleet rebalancing has long been studied using centralized approaches, meaning the service provider takes decisions expected to maximize its performance and provides directions to drivers accordingly. Drivers are assumed to comply with it. However, the TNCs now generally rely on decentralized decision-making processes, with self-employed drivers determining whether or not to accept a ride or to rebalance to high-demand areas and TNCs nudging them with monetary bonuses.
Therefore, the research teams at EPFL (Lausanne, Switzerland) and Univ. Gustave Eiffel (Lyon, France) have developed and compared several distributed rebalancing strategies. The methods developed cover a large range of methodological approaches. They include:
- An auctioning-based system to dispatch idle vehicles based on the outcome of a distributed two-sided matching process;
- A hierarchical distributed control strategy to reorganize fleets both at the microscopic and macroscopic scales;
- A pricing scheme to rebalance vehicles through targeted ride-splitting.
These strategies are compared for the first time in a common case study, the city of Lyon in France. This joint analysis evidences a general increase in the number of passengers served, while some of the proposed strategies also reduce passengers waiting time.
Find the full report (D3.3) in our publication repository or following this link.