Shih-Fen Cheng

Search frictions and misallocation are common in decentralized transportation markets. Using novel trip-level data of taxis in Singapore, this paper examines the impact of real-time demand information at airport terminals on search frictions. The information reduces taxi supply misallocation, increasing deadheading speed by 16.3% and decreasing deadheading time by 10.77%, benefiting both passengers and drivers. It raises daily earnings by $3.70 USD and adds 6.2 minutes of operational time per airport-trip taxi. Spatial spillovers are primarily observed among drivers in adjacent districts. Taxis from the Budget Terminal and drivers with fewer prior airport pickups benefit more from this information.

Freight forwarding plays a crucial role in facilitating global trade and logistics. However, as the freight forwarding market is extremely fragmented, freight forwarders often face the issue of not being able to fill the available shipping capacity. This recurrent issue motivates the creation of various freight forwarding networks that aim at exchanging capacities and demands so that the resource utilization of individual freight forwarders can be maximized. In this paper, we focus on how to design such a collaborative network based on collaborative game theory, with the Shapley value representing a fair scheme for profit sharing. Noting that the exact computation of Shapley values is intractable for large-scale real-world scenarios, we incorporate the observation that collaboration among two forwarders is only possible if their service routes and demands overlap. This leads to a new class of collaborative games called the Locally Collaborative Games (LCGs), where agents can only collaborate with their neighbors. We propose an efficient approach to compute Shapley values for LCGs, and numerically demonstrate that our approach significantly outperforms the state-of-the-art approach for a wide variety of network structures.

Complex planning and scheduling problems have long been solved using various optimization or heuristic approaches. In recent years, imitation learning that aims to learn from expert demonstrations has been proposed as a viable alternative to solving these problems. Generally speaking, imitation learning is designed to learn either the reward (or preference) model or directly the behavioral policy by observing the behavior of an expert. Existing work in imitation learning and inverse reinforcement learning has focused on imitation primarily in unconstrained settings (e.g., no limit on fuel consumed by the vehicle). However, in many real-world domains, the behavior of an expert is governed not only by reward (or preference) but also by constraints. For instance, decisions on self-driving delivery vehicles are dependent not only on the route preferences/rewards (depending on past demand data) but also on the fuel in the vehicle and the time available. In such problems, imitation learning is challenging as decisions are not only dictated by the reward model but are also dependent on a cost-constrained model. In this paper, we provide multiple methods that match expert distributions in the presence of trajectory cost constraints through (a) Lagrangian-based method; (b) Meta-gradients to find a good trade-off between expected return and minimizing constraint violation; and (c) Cost-violation-based alternating gradient. We empirically show that leading imitation learning approaches imitate cost-constrained behaviors poorly and our meta-gradient-based approach achieves the best performance.

This paper investigates the impact of neighborhood retail amenities on taxi trip behavior. We exploit a natural experiment in Singapore where the government introduced a new policy that allows retail shops to operate in residential neighborhoods. We find that the introduction of retail amenities in residential neighborhoods leads to a significant increase in taxi trips. This increase is driven by the increase in the number of trips to and from the retail amenities. We also find that the increase in taxi trips is more pronounced during the weekends and evenings. Our findings suggest that the introduction of retail amenities in residential neighborhoods can have a significant impact on taxi trip behavior.

We study the role of ride-hailing surge factors on the allocative efficiency of taxis by combining a reduced-form estimation with structural analyses using machine-learning-based demand predictions. Where other research study the effect of entry on incumbent taxis, we use higher frequency granular data to study how location-time-specific surge factors affect taxi bookings to bound the effect of customer decisions while accounting for various confounding variables. We find that even in a unique market like Singapore, where incumbent taxi companies have app-based booking systems similar to those from ride-hailing companies like Uber, the estimated upper bound on the cross-platform substitution between ride-hailing services and taxi bookings is only 0.26. On the other hand, we show that incorporating surge price factor improves the precision of demand prediction by 12% to 15%. Our structural analyses based on a driver guidance system finds this improved accuracy in demand prediction reduces drivers’ vacant roaming times by 9.4% and increases the average number of trips per taxi by 2.6%, suggesting the price information is valuable across platforms, even if elasticities are low.

The arrangement of last-mile services is playing an increasingly important role in making public transport more accessible. We study the use of ridesharing in satisfying last-mile demands with the assumption that demands are uncertain and come in batches. The most important contribution of our paper is a two-level Markov decision process framework that is capable of generating a vehicle-dispatching policy for the aforementioned service. We introduce state summarization, representative states, and sample-based cost estimation as major approximation techniques in making our approach scalable. We show that our approach converges and solution quality improves as sample size increases. We also apply our approach to a series of case studies derived from a real-world public transport data set in Singapore. By examining three distinctive demand profiles, we show that our approach performs best when the distribution is less uniform and the planning area is large. We also demonstrate that a parallel implementation can further improve the performance of our solution approach.