Conference Agenda

Crowdsourced delivery offers a compelling alternative to traditional delivery services, with benefits including reduced costs, faster delivery times, greater adaptability, and contributions to sustainable urban logistics. However, the compensation that a crowdsourced platform offers to gig workers significantly affects their acceptance probability for each delivery request. Therefore, the success of a platform that utilizes crowdsourced delivery relies on finding a pricing policy that strikes a balance between creating attractive offers for gig workers and ensuring profitability. In this work, we study a dynamic pricing problem from the perspective of a central operator responsible for determining request-specific compensation for gig workers. We examine a discrete-time framework where delivery requests and gig workers arrive stochastically. Each gig worker is willing to serve one request at a time for a fee. The operator aims to find a pricing policy that maximizes the total expected reward from servicing customer requests within the time horizon. In this context, we employ the Multinomial Logit model to represent the acceptance probabilities of drivers. As a result, we can derive an exact solution that utilizes the post-decision states, i.e., the intermediate states immediately after a driver has decided on which delivery request to accept but before any additional stochastic information is revealed to the system. Subsequently, we integrate this solution into an approximate dynamic programming algorithm. We compare our algorithm against benchmark algorithms, including formula-based policies and the upper bound provided by the full information linear programming solution, and show that it is superior to existing algorithms.

The widespread adoption of digital distribution channels both enables and forces more and more logistical service providers to manage booking processes actively to maintain competitiveness. As a result, their operational planning is no longer limited to solving vehicle routing problems. Instead, demand management and subsequent vehicle routing problems are integrated to steer the booking process with the aim of optimizing the downstream fulfillment operations. The resulting integrated demand management and vehicle routing problems (i-DMVRPs) can be modeled as Markov decision process and, theoretically, solved via the well-known Bellman equation. Unfortunately, the Bellman equation is intractable for industry-sized instances. Thus, in the literature, i-DMVRPs are often addressed via opportunity cost approximation approaches. However, the overall performance of the respective approaches largely varies between different instance structures. Furthermore, to the best of our knowledge, there is neither a structured procedure to analyze the corresponding root causes nor general guidelines on when to apply which class of approximation approach. In this work, we address this gap by proposing a structured method to analyze, explain and compare the performance impact of different opportunity cost approximation-based solution approaches for i-DMVRPs. Further, we identify common patterns in approximation errors and derive general guidelines for an informed algorithm development process.

Amid the rapid growth of online retail, last-mile delivery faces significant challenges, including the cost-effective delivery of goods to all customers. Accordingly, the development and improvement of innovative approaches thrive in current research. Our work contributes to this stream by applying dynamic pricing techniques to effectively model the possible involvement of the crowd in fulfilling delivery tasks. The use of occasional drivers (ODs) as a viable, cost-effective alternative to traditional dedicated drivers (DDs) prompts the necessity to focus on the inherent challenge posed by the uncertainty of ODs' arrival times and willingness to perform deliveries.

We propose a dynamic programming framework that determines specific compensation for each delivery task to engage the crowd. This model, resembling a reversed form of dynamic pricing, incorporates ODs’ decision-making by considering their unknown destinations and potential detours when accepting a delivery task. Our approach addresses the dynamic and stochastic nature of OD availability and decision-making.

We analytically solve a one-dimensional optimization problem, where all delivery locations are on a straight line, and identify key properties. These insights are then extended to the more complex two-dimensional case. We present our solution method for developing an effective compensation scheme, demonstrating its potential to improve last-mile delivery efficiency.