Abstract
This paper aims to combine both economic and network user equilibrium for ride-sourcing and ride-pooling services, while endogenously optimizing the pooling sequence of two origin-destination (OD) pairs. With the growing popularity of ride-sourcing and ride-pooling services provided by transportation network companies (TNC), there lacks a theoretical network equilibrium model that accounts for the emerging ride-pooling service, due to the challenge in enumerating all possible combinations of OD pairs pooling and sequencing. This paper proposes a unified network equilibrium framework that integrates three modules, including travelers' modal choice between e-pooling and e-solo services, e-platforms' decision on vehicle dispatching and driver-passenger matching, and network congestion. To facilitate the representation of vehicle and passenger OD flows and pooling options, a layered OD graph is created encompassing ride-sourcing and ride-pooling services over origins and destination nodes. Numerical examples are performed on both small and large road networks to demonstrate the efficiency of our model. The proposed equilibrium framework can efficiently assist policymakers and urban planners to evaluate the impact of TNCs on traffic congestion, and also help TNCs with pricing and fleet sizing optimization.
Introduction
Transportation Network Companies (TNCs) have revolutionized urban mobility, but they present a complex challenge for urban management. While providing convenient transit, they often increase "deadhead" miles—drivers cruising without passengers—and exacerbate traffic congestion. Ride-pooling (e.g., UberPool) is the industry's primary solution to reduce vehicle counts, yet it currently accounts for only a fraction of total trips.
Modeling these systems holistically has historically been a combinatorial nightmare. As the number of passengers increases, the possible sequences for pick-ups and drop-offs grow quadratically ($O(N^2)$). Previous research often simplified passenger choices or ignored physical traffic congestion. This paper bridges the gap between economic supply-demand models and physical traffic assignment models by introducing a unified equilibrium framework.
Methodology
The proposed framework integrates three interacting modules to capture the full scope of urban e-hailing dynamics:
- E-Platform Operation: A minimum-cost circulation flow formulation solves how platforms dispatch idle vehicles and match them with passengers.
- Network Congestion: A standard traffic assignment problem (TAP) calculates how vehicle flows (solo, pooled, and empty) contribute to road delays.
- Customer Mode Choice: A utility-based model where travelers choose between "e-solo" and "e-pooling" based on fares, wait times, and the "inconvenience cost" of sharing a ride.
The Layered Origin-Destination (OD) Graph
The core innovation of this research is the Layered OD Graph. Instead of enumerating every possible pooling combination, the authors developed a four-layer graph structure: Solo Passenger, Solo Vehicle, Pooling Vehicle, and Pooling Passenger. This architecture allows the model to treat complex pooling sequences as simple paths through the layers, transforming a discrete sequencing problem into a continuous flow problem. This approach is computationally efficient enough to be applied to large-scale city networks.
The entire system is solved as a Nonlinear Complementarity Problem (NCP), ensuring that at equilibrium, no passenger, driver, or platform can improve their outcome by unilaterally changing their strategy.
Key Insights and Results
The model was validated using both a 3-node toy network and the standard Sioux Falls benchmark network (24 nodes, 76 links). Several key findings emerged:
- Computational Scalability: The layered OD graph approach successfully bypassed the $O(N^2)$ complexity bottleneck, proving viable for realistic urban scenarios.
- The Congestion Feedback Loop: The research identifies a "vicious cycle" in pooling. High congestion increases the time required to pick up a second passenger. This increases wait times, driving passengers toward solo rides, which in turn further increases congestion.
- Price Sensitivity: Results suggest pooling is highly sensitive to price. In many scenarios, passengers required discounts of 75% or more to tolerate the detours and wait times inherent in pooled services.
- Emergent Surge Pricing: The model demonstrates that "surge pricing" emerges mathematically as a multiplier when vehicle supply cannot meet demand at specific locations, rather than being a purely arbitrary business tactic.
Conclusion and Policy Implications
The study demonstrates that simply increasing the number of TNC vehicles in a city may be counterproductive, potentially increasing deadhead miles and congestion. Instead, urban policies should focus on breaking the congestion feedback loop by subsidizing pooling or implementing congestion pricing for solo rides.
Future work will look to expand this model to include competition between multiple TNC platforms and dynamic, real-time dispatching scenarios, as the current model focuses on long-term planning (steady-state) equilibrium.
Summary updated