Large-scale Online Ridesharing: The Effect of Assignment Optimality on System Performance

TL;DR

This work tackles large-scale MoD ridesharing by applying the Vehicle Group Assignment (VGA) method to compute optimal passenger–vehicle assignments and routes under a QoS limit, demonstrating substantial improvements over insertion heuristics and non-ridesharing baselines. The authors scale VGA to scenarios with more than $21{,}000$ active requests and $11{,}000$ vehicles using a station-based, demand-driven MoD model for Prague and compare multiple configurations, including resource-constrained VGA variants. Key findings show that optimal VGA can reduce total fleet distance by over $57\%$ relative to no ridesharing and by about $20\%$ relative to IH, while maintaining passenger delays near $4$ minutes; sensitivity analyses reveal limits on batch length and delay for which optimality can be certified. The study highlights significant congestion reduction and fleet-efficiency gains from optimal ridesharing, while also outlining practical trade-offs between computation time and solution quality, and charting directions for scalable, multi-objective MoD design.

Abstract

Mobility-on-demand (MoD) systems consist of a fleet of shared vehicles that can be hailed for one-way point-to-point trips. The total distance driven by the vehicles and the fleet size can be reduced by employing ridesharing, i.e., by assigning multiple passengers to one vehicle. However, finding the optimal passenger-vehicle assignment in an MoD system is a hard combinatorial problem. In this work, we demonstrate how the VGA method, a recently proposed systematic method for ridesharing, can be used to compute the optimal passenger-vehicle assignments and corresponding vehicle routes in a massive-scale MoD system. In contrast to existing works, we solve all passenger-vehicle assignment problems to optimality, regularly dealing with instances containing thousands of vehicles and passengers. Moreover, to examine the impact of using optimal ridesharing assignments, we compare the performance of an MoD system that uses optimal assignments against an MoD system that uses assignments computed using insertion heuristic and against an MoD system that uses no ridesharing. We found that the system that uses optimal ridesharing assignments subject to the maximum travel delay of 4 minutes reduces the vehicle distance driven by 57 % compared to an MoD system without ridesharing. Furthermore, we found that the optimal assignments result in a 20 % reduction in vehicle distance driven and 5 % lower average passenger travel delay compared to a system that uses insertion heuristic.

Figures (18)

• Figure 1: Two example trips from the generated demand. The filled circles represent activities, while the arrows represent trips between those activities. Next to each activity, we can see the person and activity IDs in the format person_id-activity_id. The activities corresponding to these IDs can be found in Table \ref{['fig:activities']}.
• Figure 2: Demand for personal vehicle traffic in Prague. The start positions of all vehicle trips are discretized to squares of 200.0 square meters. Darker color translates to higher demand, and the color bar has a logarithmic scale.
• Figure 3: Histogram of fastest path travel times for each trip.
• Figure 4: MoD system stations in the city of Prague. There are 73 stations in total, shown as red circles.
• Figure 5: Example of the VGA method assigning three passengers to two vehicles. In Figure \ref{['fig:vga_eample1']}, we show all possible request groups for each vehicle. The lines between the request (left) and the group (middle) denote the membership in the group. The lines between the groups and vehicles denote feasible group assignments. In Figure \ref{['fig:vga_eample2']}, the final assignment between vehicles and groups is shown (bold lines).