Preprint and Working Papers
Recipient-Dependent Last-Mile Delivery Routing with Autonomous Vehicle Applications
status: Major Revision at Transportation Science Journal
Bahar D Viniche, Mehdi Noruinejad, Opher Baron, Oded Berman
Last-mile delivery contributes to a drastic 28% of the total cost of shipping. Recent technological advance- ments enable recipients to participate in these deliveries, relieving the cost of this final leg of the supply chain. We present models of recipient-dependent routing policies in last-mile logistics. The policies are not limited to but are motivated by the future applications of autonomous vehicles in smart cities and their role in enabling recipient-dependent deliveries as they relinquish the need for drivers. The policies are based on scenarios where recipients use their AVs to pick up from (i) a central depot, (ii) a hub located by the logistics firm close to them (hybrid), and (iii) a hub and deliver to other nearby recipients (crowdsourcing). We compare the policies with status quo truck delivery and investigate the potential cost savings. The anal- ysis shows a robust dominance space pattern against key operational parameters. In particular, (i) status quo truck routing is the currently preferred delivery policy, (ii) the hybrid and crowdsourcing policies are a combination of AV and status quo truck policies in terms of dominance space, and (iii) recipient-dependent routing policies dominate the status quo truck policy as the number of recipients increases. We validate the insights from the stylized model with a case study of Walmart in Toronto.
Tactical Fleet Planning in Drone Enabled Deliveries
status: To be submitted soon to Production and Operation Management Journal
Bahar D Viniche, Mehdi Noruinejad, Opher Baron, Oded Berman
Drones offer substantial value to last-mile logistics in both road-congested urban areas and rural delivery locations by avoiding congested routes and follow aerial pathways at higher speeds. Strategically, integrating drones into an existing fleet requires a balanced assessment of reduced delivery times against the additional acquisition costs. We study the optimal fleet composition problem for multi-truck multi-drone parcel delivery by proposing a parametric design and providing strategic analyses. This research present models of drone-assisted delivery strategies where trucks travel from the central depot to designated hubs from which drones are launched and retrieved. This research bridges the gap of incorporating routing-based operational level constraints where the optimized routing strategy and fleet composition are introduced based on constraints such as drone flight range, payload capacities of trucks and drones, and their respective speeds. We introduce multi-launching of drones, where each truck can launch more than one drone at each launch hub, and drone re-launching, which allows a drone to return to the truck after finalizing a delivery, resupply, and launch for the next delivery, provided it remains within the flight range. We consider the synchronization of truck movements in terms of both minimizing the delivery time and a scheduling constraint to ensure the timely coordination between drones and their associated truck. The analysis reveals three scenarios in which the delivery efficiency is constrained by i) the truck’s capacity, ii) the drone’s range, or iii) the timely coordination between trucks and drones. These results highlight technological limitations that hinder the efficiency of drone-enabled delivery, providing managerial insights about components that require improvement and when certain constraints become binding based on the current technology.
Evaluating Drone-Delivery Efficiency in Different Urban Settings Using GNNs
status: Work in Progress
Bahar D Viniche, Mehdi Noruinejad, Opher Baron
Drones have the potential to shorten delivery times and alleviate challenges such as parking in dense urban areas or ensuring timely delivery to remote locations. Although drones promise faster and more cost-effective delivery, their efficiency varies across different urban settings, including city centers and suburban areas. We use ML techniques, in particular Graph Neural Network (GNN) to evaluate drone performance across these settings to address this. GNN is powerful framework for applying deep learning to graph-structured data, such as city networks. We use city structure data, delivery data, and socioeconomic data for each urban setting. We use OpenStreetMap for extracting urban structural data and constructing city representative networks. We measure drone delivery efficiency based on travel time and distance reduction data, for which we developed and leverage the “Drone Sidekick Tool”. This interactive tool allows the user to choose the location of the central depot and customers. It collects data on travel distance, travel time, and environmental impacts of drone delivery implementation. Combining this tool with the algorithmic approach, this research offers comprehensive insights into how drone delivery can transform urban logistics and supports the development of more efficient and adaptable delivery solutions.
Equitable Territory Planning for Last-Mile Routing with Temporal Flexibility
status: Work in Progress
Bahar D Viniche, Ahana Malhorta, Elkafi Hassini, Mehdi Nourinejad
Optimizing last-mile logistics is a multifaceted challenge that significantly impacts both distributors and recipients. Traditional routing optimization focuses on minimizing time and cost, often overlooking sustainability and equity. This study introduces a comperhensive routing policy for last-mile delivery that balances these criteria with route flexibility and consistency of service territory for each driver to improve driver performance and customer satisfaction. We develop a strategic approach for territory planning under capacity constraint, and stochastic demand with known demand point locations. Considering workload balance prevents driver burnout and maintains high service quality. The model introduces the concept of “flex zones” to adjust driver workloads based on daily delivery demands and the concept of “core zones” to incorporate driver familiarity with territories. Integrating core and flex zones within each delivery territory minimizes delivery time while maintaining route flexibility and familiarity. This routing strategy improves overall delivery efficiency. The proposed framework is validated through detailed numerical simulation, demonstrating its potential to create a more adaptive and resilient last-mile delivery system.