Route efficiency implications of time windows and vehicle capacities in first- and last-mile logistics
This paper investigates the route efficiency effects of combining first-mile pickup and last-mile delivery operations with realistic constraints. It proposes closed-form adjustment factors, derived using a hybrid data analysis approach, to capture the impact of time window and vehicle capacity constraints on efficiency. These factors help in optimizing strategic design and operational planning of industrial-scale distribution networks amid increasing e-commerce volumes and logistics cost pressures.
Crowdsourced on-demand food delivery: an order batching and assignment algorithm
This paper presents an order batching and assignment algorithm for online food delivery platforms, using a graph-based approach and decomposing the problem into tractable sub-problems. The study integrates advanced policies, including 'Insertion' and 'Swap' policies, and develops an agent-based simulation framework. Results show considerable improvement in solution quality across different demand and supply patterns based on a real-world case study.
This paper introduces a model for optimal placement of loading bays in cities, considering delivery demand, bay capacity, and driver walking times. The model allows selective non-allocation of bays if land-use costs exceed traffic disruption costs, considering traffic, parking probability, and a policymaker-specified sensitivity factor.
This paper presents an optimization model integrating collection-and-delivery points (CDPs) into the design of an efficient retail distribution network, accounting for changes in demand density, failed deliveries, and return pickups. The model is scalable and tested on a Brazilian retailer's network, demonstrating reduced last-mile and overall distribution costs through CDPs.
This report analyzes legal, regulatory, and societal challenges facing unmanned aerial vehicles (UAVs) used for last-mile delivery in the U.S. It explores current restrictions, areas of uncertainty, and societal barriers, providing insight into potential limitations for operators and impacts on network planning and strategy.
The 2021 Amazon Last Mile Routing Research Challenge, supported by MIT, used real operational data to solve routing issues. The provided dataset included details from 9,184 historical routes from 2018, covering five U.S. metropolitan areas, while ensuring anonymity. This is the first large, public, real-world operational routing dataset.
The paper presents a two-stage stochastic program for designing two-echelon last-mile delivery networks amidst demand uncertainty. The approach divides the problem into strategic (facility location) and operational (daily distribution) decisions. Using a case study from New York City, the model provides insights on transportation modes, facility location, and outsourcing impacts.
This study proposes a metamodel simulation-based optimization approach to strategically design last-mile distribution networks for online and omnichannel retailers with tight delivery deadlines. It outperforms contemporary approaches and traditional methods, offering superior cost performance and consistency. The study shows that facility congestion impacts late deliveries and consolidation potential, especially under tighter deadlines.
This paper compares three truck-and-drone delivery models for package delivery, focusing on their synchronization level. It formulates associated routing problems and calculates maximum possible savings compared to truck-only routes. The study finds that higher synchronization significantly reduces customer waiting times, with potential reductions over 60% under certain conditions.
This paper presents a stakeholder participation approach that seeks cognitive consensus in developing and prioritizing urban freight logistics policies. Using the Analytic Hierarchy Process within an action research framework, it illustrates the importance of revealing and sharing policy assumptions in local contexts. It shows that structured stakeholder involvement leads to shifts in policy priorities.
This paper proposes an integrated framework for designing three-tiered multi-modal networks in an omni-channel environment, considering diversified customer demand and product-exchange options. The model extends route cost estimation methods for large-scale demands and optimizes network configurations. Numerical experiments and a São Paulo case study highlight the economic benefit of this integrated approach.
This book chapter explores interactive visualization and human-in-the-loop optimization to enhance decision making in Supply Chain Design. It combines data, analytical models, and expert knowledge through innovative human-machine interfaces. Drawing from macro-cognitive theory, it proposes a framework for group decision-making processes, and presents two case studies using this approach in multinational companies.
This study analyzes the optimal setup for in-store fulfillment of online orders for omni-channel retailers, utilizing a simulation-based approach. Applied to a real-world case, it determines optimal order batching time, picking process time, number of pickers and packers, and associated performance measures. The analysis results provide broad managerial implications.
This paper presents a structured framework for classifying drone-based delivery systems and their routing problems, along with a comprehensive literature review. It identifies research gaps and underrepresented applications, such as food and mail deliveries. The review reveals a focus on single-truck, single-drone models and a lack of attention to resupply multi-modal models.
This paper reviews the literature on on-demand food delivery (ODFD) services, highlighting the need for an integrated, ecosystem-based view, and focusing on restaurant operations. It also suggests further research in human resource management and logistics of ODFD systems, specifically in intervention/regulation and distribution network/batching. The review provides insights for both academics and practitioners.
The paper proposes a demand forecasting approach for meal delivery platforms like Uber Eats, blending classical forecasting and machine learning methods. The study reveals that an exponential smoothing method trained on past demand data yields optimal accuracy with a two-month history. For shorter history, machine learning provides better predictions.
The paper proposes a large-scale stochastic mixed integer linear programming model to strategically design efficient urban distribution networks in emerging economies, taking into account demand uncertainty and physical distribution flexibility. Applied to real-world case studies from Coca-Cola Femsa's operations in Colombia, results show the stochastic design outperforms deterministic design in terms of cost and performance risk. However, in some deterministic scenarios, operational flexibility might lower network performance.
This report evaluates the integration of a truck-and-drone delivery system in urban environments, assessing the current regulatory environment, particularly in the US, and discussing the social, environmental, and operational implications of its large-scale deployment. It concludes with a proposal for an adaptive regulatory framework, considering various stakeholders, to guide the technology's responsible development and deployment.
This paper analyzes local network circuity, revealing significant heterogeneities explained by road network properties, using traffic datasets and machine learning methods. Using São Paulo as a primary case study and comparing it to seven other urban areas, the study finds that real trip distances are roughly twice the distances predicted by the L2 norm.
This study expands the Traveling Repairman Problem (TRP) by introducing a single truck launching drones multiple times at each stop. The mathematical model and a hybrid Tabu Search-Simulated Annealing algorithm indicate that this approach could significantly reduce customer waiting times across various parameters compared to traditional delivery models.
This study introduces a mathematical model and heuristic approach for optimizing delivery routes in a synchronized truck and Unmanned Aerial Vehicle (UAV) system. By using a Mixed-Integer Linear Programming model and a Truck and Drone Routing Algorithm, the study aims to minimize customer waiting time, determining optimal routes and dispatch points. Applied to a real-world case in São Paulo, results show significant potential for reducing customer waiting times.
This paper presents an integrated framework for characterizing urban last-mile e-commerce distribution strategies in both mature and emerging markets. Through literature review and case studies, it compares different strategies and identifies variables impacting network design decisions, filling a gap in literature regarding strategies tailored to emerging markets.
This paper proposes a machine-learning approach for inferring customer constraints from transactional data to optimize routing in supply chain and logistics. Using a probabilistic directed graphical model and a Metropolis-Hastingswithin-Gibbs sampling algorithm for inference, the method outperforms simple counting of occurrences, suggesting potential for future research combining machine learning and routing problems.
This paper analyzes the efficiency trade-offs in urban distribution systems from integrating first-mile pickup and last-mile delivery operations. The authors propose adjustment factors to improve continuum approximation-based route length estimations, showing potential efficiency gains of up to 30% from integration. Applied to a real-world case in India, the method suggests reduced traffic and emissions by up to 16%, and improved fleet utilization and cost.
This paper reviews revenue management in last-mile delivery, particularly attended home deliveries of groceries, and suggests future research avenues. It presents a topological classification of last-mile delivery characteristics, identifies extensions based on current industry trends, and discusses how existing models can be adapted to these new problems.
The study presents a method for integrating collection-and-delivery points (CDPs) in multi-echelon distribution network design. A non-linear optimization model is formulated, considering CDP location decisions and changing demand patterns. Extended routing cost approximation formulae are developed for large-scale problems. The model is applied to a case study involving a major Brazilian e-commerce firm.
This paper applies a Complex Adaptive System (CAS) perspective to urban logistics decision-making, addressing challenges like stakeholder objectives, unclear problem ownership, and data scarcity. An integrative framework is proposed to improve urban logistics policies. The CAS perspective is applied to a case study, demonstrating its potential in real-world urban logistics policy-making scenarios.
This paper introduces a data-driven extension to continuum approximation methods for predicting urban route distances, improving accuracy by incorporating road network circuity. Compared to traditional methods, the extension reduces mean absolute percentage error by 26 points, producing estimates within 5-15% of near-optimal solutions. This significantly enhances the real-world validity of large-scale logistics system design and planning.
This study uses large-scale urban sensing data, specifically GPS vehicle traces, to understand and characterize urban logistics in São Paulo, Brazil. The approach analyses both company-specific and mixed-use general vehicle fleets data, yielding insights into spatial distribution, freight traffic flows, stop times, and speed patterns. This method can be extended to other geographic contexts and similar datasets.
This paper reviews case studies and surveys to examine how retailers manage last-mile distribution in an omnichannel environment. It proposes a typology of four ideal last-mile supply networks differentiated by delivery speed and product variety. It also provides guidelines for retailers to reconfigure their last-mile distribution networks.
The study explores the use of motorized cargo tricycles alongside traditional trucks in a mobile depot-based system for urban freight delivery. It proposes a method to assess the impact on service level, emissions, and delivery cost. The results suggest this strategy can significantly reduce greenhouse gas emissions and local pollutants, and yield cost advantages in areas with low average delivery drop sizes.
This study presents a decision model to assess the cost benefits of merging urban mail and parcel delivery networks for La Poste, the French national postal operator. Using data from Nantes, France, the model finds that merging these networks could yield around a 3% cost reduction. This insight led to organizational changes at La Poste.
This paper presents a mixed-integer linear programming model to solve a two-echelon capacitated location-routing problem for urban logistics services. The model aims to optimize facility network and vehicle fleet design for postal operators. We propose an optimal routing cost estimation formula and an optimization heuristic, enabling efficient solutions for large-scale problems.
Developing a flight plan for drone-based parcel delivery
Although the regulatory framework is central to the industry’s future, it remains unclear. But, where there are problems there are also opportunities, especially in an industry that is developing innovative technologies and new operational and business models.