One of the key decisions for an on-demand service platform is to plan capacity to meet uncertain demand. This problem is also compounded by the operating environment and multiple stakeholder perspectives. For example, capacity is typically determined not only by multiple supply sources but also by the platform’s compensation scheme, as this affects labor pool availability. In addition, since on-demand platforms do not service demand using permanent (e.g., full-time) employees, it is likely that the employee pool is heterogeneous in their income preferences. In this paper, we analytically characterize the capacity planning problem for an e-hailing platform offering transportation service to customers (such as Uber and Lyft) using independent agents (or drivers). In the presence of uncertain demand, the unique features incorporated into our analysis include sources of supply (single/dual), driver absenteeism rates, platform compensation schemes, labor pool constraints, and heterogeneity in drivers’ income-earning orientation. Interestingly, one of our major findings is that labor pool constraints determine the types of drivers that the platform should recruit. In the absence of such constraints, the platform should use only “unreliable” drivers, whereas both reliable and unreliable drivers should be employed when the labor pool is constrained. From a platform perspective, a lower compensation fraction should be offered under a post-paid scheme than under a pre-paid compensation scheme. The model and its results are validated using empirical data from different markets. A sensitivity analysis is performed to assess the robustness of this approach across various demand, payment, and driver-type scenarios.

Road transportation via trucks is a dominant mode for long-haul freight transport across countries. However, due to their significant dependence on fossil fuels, trucks are a large contributor to carbon emissions. Hence, new technology-driven solutions such as truck platoons are gaining momentum. While platoons promise to reduce fuel costs and emissions, they may increase transportation time due to additional coordination delays, such as the time required for platoon formation. In this research, we examine the performance trade-offs between platoon fuel savings and excess delay costs resulting from waiting for platoon formation among three platoon formation strategies: intermittent, continuous, and opportunistic. We develop a novel Closed Queuing Network model that captures the dynamics of platoons, as well as the stochasticity in truck travel times, and provides realistic estimates of platoon wait times and vehicle throughput. The platoon formation delays and size-dependent travel times are modeled using merging and load-dependent nodes, respectively, and analyzed through a continuous-time Markov chain. Our study provides key insights into the impact of increasing platoon size on performance measures, including system throughput and mean waiting time. With platooning, the network throughput capacity is reduced; however, fuel savings are realized. For a given network topology, we can identify an optimal platoon formation strategy that maximizes the throughput and fuel efficiency, while simultaneously minimizing vehicle waiting costs.

Motivated by the recent exits of automotive original equipment manufacturers (OEMs) from various markets, we study a stochastic capacity allocation problem in which the capacity allocation decisions of a dominant supplier influence both market demand and the exit risk of an OEM. For example, a shortage of critical components can jeopardize the OEM’s profitability, ultimately increasing its risk of market exit. We develop a stochastic capacity allocation model where the buyer (i.e., OEM) faces service-dependent demand and market exit risk. The customer demand is modeled as a function of the OEM’s market goodwill, which evolves based on the component supply from the supplier. We show that the supplier’s optimal capacity allocation policy follows a goodwill-dependent threshold policy characterized by two control limits, which depend on the OEM’s market goodwill, risk tolerance, and profit objectives. Our analysis yields several key insights. First, even when component supply is more critical for a fragile OEM, the supplier may sometimes allocate less capacity to the fragile OEM than to a non-fragile one. Second, when the OEM faces service-dependent demand, the supplier strategically allocates more capacity than in scenarios without service-dependent demand. Finally, we observe that as customers emphasize recent experiences, the optimal capacity allocation increases. This heightened sensitivity necessitates more careful handling by the OEM, prompting the supplier to ensure a more reliable supply of components. The insights from our study provide suppliers of critical components with valuable strategies for managing production for OEMs that are facing service-dependent demand.

E-commerce order fulfillment is increasingly disrupted by natural events such as pandemics, hurricanes, and floods. This study investigates order assignment decisions considering warehouse disruption risk, order-class priority, and shipping costs. We develop a stochastic dynamic programming model for the order assignment problem. Our analysis reveals a switching-curve policy for order assignment. We find that disruption risk significantly affects the order assignment decision, with optimal switching thresholds decreasing as the disruption rate increases. To efficiently compute these thresholds, we develop three index-based heuristic policies. Among them, our improvement heuristic achieves an average optimality gap of 7.21%, outperforming the myopic policy (8.48%) and the least shipping cost heuristic (14.17%). Through a comprehensive numerical study, we uncover several important insights. Disruption and recovery rates have nonlinear effects on order fulfillment costs. Specifically, while investing in mechanisms to enhance recovery speed is beneficial, the gains become progressively smaller as recovery becomes faster. Additionally, shared order-processing capacity at warehouses with class-wise priority can prove a more effective strategy than maintaining dedicated capacities for each order class. This research provides actionable strategies for managing e-commerce fulfillment under warehouse disruption risks, enhancing operational efficiency and cost management.

The rapid growth of e-commerce has increased the demand for efficient order picking systems in large warehouses. To improve throughput performance, many facilities deploy autonomous mobile robots (AMRs) to assist human pickers. Warehouse throughput critically depends on the choice of human-robot collaboration policy. This study focuses on two popular policies: the swarm policy, in which pickers switch between AMRs while picking, and the system-directed policy, in which a picker completes an order with a single AMR. An analytical framework is developed to evaluate these policies. We model the swarm policy as a closed queuing network with a synchronization station, and we derive closed-form expressions for its steady-state probabilities and throughput given load-dependent service rates. The service rates of the network nodes are estimated by Monte Carlo simulation, accounting for stochastic travel times, varying order sizes, item allocation strategies, matching rules, and warehouse layouts. The analytical predictions are validated against detailed discrete-event simulations, with average relative errors below 2% in [Formula: see text] instances. The results indicate that the swarm policy generally provides higher throughput than the system-directed policy, with gains increasing in the AMR-to-picker count and speed ratios. The system-directed policy is more effective when AMR and picker speeds are similar, the orders are large, and there is a limited number of AMRs. Managerial insights are provided to guide policy choice. Funding: This research is part of the Sharehouse Project, which was cofinanced and supported by the Dutch Research Council NWO, the Dutch Ministry of I&W, the Taskforce for Applied Research SIA, the Dutch Topsector Logistics, and TKI Dinalog [Project 439.18.452]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/trsc.2024.0969 .

Membrane potential (MP) changes can provide a simple readout of bacterial functional and metabolic state or stress levels. While several optical methods exist for measuring fast changes in MP in excitable cells, there is a dearth of such methods for absolute and precise measurements of steady-state MPs in bacterial cells. Conventional electrode-based methods for the measurement of MP are not suitable for calibrating optical methods in small bacterial cells. While optical measurement based on Nernstian indicators have been successfully used, they do not provide absolute or precise quantification of MP or its changes. We present a novel, calibrated MP recording approach to address this gap. In this study, we used a fluorescence lifetime-based approach to obtain a single-cell-resolved distribution of the membrane potential and its changes upon extracellular chemical perturbation in a population of bacterial cells for the first time. Our method is based on 1) a unique VoltageFluor (VF) optical transducer, whose fluorescence lifetime varies as a function of MP via photoinduced electron transfer and 2) a quantitative phasor-FLIM analysis for high-throughput readout. This method allows MP changes to be easily visualized, recorded and quantified. By artificially modulating potassium concentration gradients across the membrane using an ionophore, we have obtained a Bacillus subtilis-specific MP versus VF lifetime calibration and estimated the MP for unperturbed B. subtilis cells to be -65 mV (in minimal salts glycerol glutamate [MSgg]), -127 mV (in M9), and that for chemically depolarized cells as -14 mV (in MSgg). We observed a population-level MP heterogeneity of 6-10 mV indicating a considerable degree of diversity of physiological and metabolic states among individual cells. Our work paves the way for deeper insights into bacterial electrophysiology and bioelectricity research.

Robotic sorting systems (RSSs) use mobile robots to sort items by destination. An RSS pairs high accuracy and flexible capacity sorting with the advantages of a flexible layout. This is why several express parcel and e-commerce retail companies, who face heavy demand fluctuations, have implemented these systems. To cope with fluctuating demand, temporal robot congestion, and high sorting speed requirements, workload balancing strategies such as dynamic robot routing and destination reassignment may be of benefit. We investigate the effect of a dynamic robot routing policy using a Markov decision process (MDP) model and dynamic destination assignment using a mixed integer programming (MIP) model. To obtain the MDP model parameters, we first model the system as a semiopen queuing network (SOQN) that accounts for robot movement dynamics and network congestion. Then, we construct the MIP model to find a destination reassignment scheme that minimizes the workload imbalance. With inputs from the SOQN and MIP models, the Markov decision process minimizes parcel waiting and postponement costs and helps to find a good heuristic robot routing policy to reduce congestion. We show that the heuristic dynamic routing policy is near optimal in small-scale systems and outperforms benchmark policies in large-scale realistic scenarios. Dynamic destination reassignment also has positive effects on the throughput capacity in highly loaded systems. Together, in our case company, they improve the throughput capacity by 35%. Simultaneously, the effect of dynamic routing exceeds that of dynamic destination reassignment, suggesting that managers should focus more on dynamic robot routing than dynamic destination reassignment to mitigate temporal congestion. Funding: This work was supported by The Fundamental Research Funds for the Central Universities [Grant WK2040000094] and the National Natural Science Foundation of China [Grants 71921001 and 72091215/72091210]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/trsc.2023.0458 .

Many e-commerce warehouses use robotic mobile fulfillment system (RMFS), where humans collaborate with robots to pick the orders. The performance of such systems depends on the joint performance of robots and humans. The performance of the workers is affected by fatigue, or the energy that it takes them to pick the items. In this paper, we study the effect of scattered storage assignment, order batching, and pod selection to minimise the total picker energy expenditure and the total robot transport distance. We introduce a mixed-integer programming formulation (called JIOPP) and introduce the NSGAII-ILS algorithm to heuristically solve it for real-world instances. Extensive numerical experiments on real-world instances show that NSGAII-ILS is competitive compared to state-of-the-art algorithms and can find Pareto solution sets that are closer to the true Pareto frontier. We evaluate the effects of batch sizes, the number of pod layers, and different pod selection policies. The results show that batching orders can save more than 35Raj of the picker's energy expenditure and more than 70Raj of the robot's transportation distance. Using the 'golden zone' layers on the pod and selecting the right pod for retrieval are important for striking a balance between worker fatigue and order picking efficiency.

Managing the performance of intralogistics operations, that is logistics operations within facilities such as manufacturing plants, order fulfillment warehouses, ports and terminals, and retail stores, is critical in fulfilling customer expectations. Traditional decision-making for intralogistics operations is based on historical data, typically collected over long-range intervals with significant processing delays. However, nowadays, Internet of Things (IoT) applications are used to gather detailed real-time data to make dynamic decisions. These new data sources provide challenges and opportunities for operations management. We provide an overview of prominent IoT technologies in four domains: Manufacturing, warehousing, ports and terminals, retail, and other emerging areas. We discuss four prominent research questions (cutting across multiple application domains) that can be addressed using new data sources, along with the methodological approach and managerial insights that may result. In particular, IoT can improve the tracking and tracing of objects, equipment, and humans and provide rapid alerts, allowing managers to make real-time decisions and improve asset use, uptime, and profitability.

Problem definition: Online food delivery (OFD) platforms have transformed the restaurant industry, prompting restaurateurs to serve both online and dine-in customers. To ease pressure on store-front kitchen capacity, restaurateurs use channel control and central kitchen strategies. Channel control determines when to close the OFD channel, while central kitchens produce semi-cooked meals to meet demand without straining the store-front kitchen capacity. To leverage these strategies, restaurateurs must manage three key decisions jointly: (i) when to close the OFD channel, (ii) how to fulfill dine-in and delivery orders, and (iii) how to replenish semi-cooked meal inventory at the store-front kitchen. Methodology/results: We study a profit-maximizing problem using a Markov decision process (MDP) model to analyze restaurateurs' joint decision-making. This model considers the spillover effect of closing the online food delivery (OFD) channel on dine-in customer demand. By examining the MDP model's structural properties, we identify optimal policies for three decisions under mild conditions and reveal the interplay between channel control and semi-cooked meal inventory, influenced by the spillover effect. A heuristic is also developed when the mild conditions are relaxed. Numerical experiments with a real-world restaurant confirm our heuristic's effectiveness (the average loss is 0.84%) and show significant value in both central kitchen and dynamic channel control policies.Managerial implications: This study reveals that closing the OFD channel, rather than opening it, benefits restaurants when the spillover effect is strong. Moreover, it shows that closing the OFD channel dynamically based on semi-cooked meal inventory, rather than at a fixed time, can significantly improve profits (up to 10%) as the spillover effect enhances. Additionally, the value of a central kitchen rises as the cost of procuring semi-cooked meals decreases.

Retailing has changed dramatically from single-channel brick-and-mortar stores to multi-channel and omnichannel retailers over the last few decades. Omnichannel retailers employ different strategies to integrate online and offline sales channels as well as order fulfillment processes. Among these strategies, the ship-from-store is the most popular and widely accepted among retailers. It enables retailers to use inventory from store locations to fulfill online demand. An omnichannel retailer with a distribution center and a retail store has to make important, interlinked decisions — (1) how much inventory to keep at the retail store, and (2) where to fulfill the online demand from and how much. In this work, we model the integrated inventory replenishment and online demand allocation decisions for an omnichannel retailer employing the ship-from-store strategy. We analyze this problem for both single-period and multi-period settings. We extend the analytical framework of the single-period problem by providing a finite-horizon Markov decision process (MDP) formulation for the multi-period problem. Our findings suggest that for a single-period setting, decentralized inventory replenishment and demand allocation system maximizes the profit of the omnichannel retailer for low values of the incentive for fulfilling the online demand through store inventory, while for sufficiently high values of the incentive, a pooled system provides the optimal profit. An increment in the discount factor has the same effect on the optimal decisions in a multi-period setting as that of salvage value in a single-period setting for a given value of the incentive for the ship-from-store strategy. We also provide several extensions (such as cross selling, endogenous and correlated demand streams) of our analytical framework for the multi-period problem.

For an e-hailing taxi operation, we analyse a driver's profit-maximising reactive strategy (to either accept or refuse a ride request) in response to the ride request broadcast by the platform. We analyse four operating modes, each of which is a combination of either of two reactive strategies: no refusal and refusal based on proximity, and either of two broadcasting methods. In an operating mode, our objective is to evaluate the expected total profit in a shift. We adopt a two-stage methodology to answer the research questions. In the first stage, we develop an agent-based simulation model to capture the effect of multiple taxis on driver's reactive strategy. Using real trip data, we find that a driver could follow a strategy of refusal based on proximity and earn approximately 25% more than the baseline no refusal strategy. In the second stage, we develop an approximate analytical model for a single taxi operation and compare the performance against the agent-based simulation model. We develop closed-form expressions of the expected total profit for each operating mode and topology of the service region. We find that our approximate analytical model provides an upper bound, and the profit deviation lies within 20% of the agent-based simulation model.

Problem definition:With a boost in digital access, online food delivery (OFD) solutions from cloud kitchens have grown significantly in recent times. This innovative business model brings unprecedented convenience and responsiveness to the customers. While OFD platforms optimize customer experience by better matching riders with customers and reducing order fulfillment times with better staffing, there is a limited understanding about the relationship between fulfillment speed and customer satisfaction ratings. Although in an e-commerce setting, studies have shown that better order fulfillment times improve sales via logistics ratings, it is unclear if such a mechanism is always true for a cloud kitchen setting. The cloud kitchen setting provides a unique context for studying a variety of brands and dish types offered to customers from a single kitchen. Different from traditional e-commerce products, where speedy delivery always brings customer delight -good food preparation and food freshness are core to customer satisfaction in OFD platforms. Methodology: We partner with a large cloud kitchen company and use large-scale transaction data from two cities to empirically examine the impact of fulfillment time on customer food ratings. Our results reveal that a delay of 10-minutes in fulfillment is associated with a 14.4% decrease in the likelihood of receiving a higher customer food rating.We find evidence supporting the idea that in an online food delivery context, perceived quality also holds importance along with timeliness, such that being too early can be detrimental. To further examine this effect, we complement our analysis with a survey experiment conducted in a controlled setting. Using insights from the empirical model as an input, we conduct post-hoc simulations to demonstrate how operational strategies such as delivery priority and workload balancing can improve fulfillment time driven customer satisfaction. Managerial implications: Our findings enable food service managers to better understand the operational consequences of fulfillment time decisions on customer satisfaction. In addition, by learning the heterogeneous effects of fulfillment time based on order characteristics and customer demographics, managers can design effective interventions to improve customer ratings.

E-commerce packages are notorious for their inefficient usage of space. More than one-quarter volume of a typical e-commerce package comprises air and filler material. The inefficient usage of space significantly reduces the transportation and distribution capacity increasing the operational costs. Therefore, designing an optimal set of packaging box sizes is crucial for improving efficiency. We present the first learning-based framework to determine the optimal packaging box sizes. In particular, we propose a three-stage optimization framework that combines unsupervised learning, reinforcement learning, and tree search to design box sizes. The package optimization problem is formulated into a sequential decision-making task called the box-sizing game. A neural network agent is then designed to play the game and learn heuristic rules to solve the problem. In addition, a tree-search operator is developed to improve the performance of the learned networks. When benchmarked with company-based optimization formulation and two alternate optimization models, we find that our ML-based approach can effectively solve large-scale problems within a stipulated time. We evaluated our model on real-world datasets supplied by a large e-commerce platform. The framework is currently adopted by a large e-commerce company across its 28 fulfillment centers, which is estimated to save the company about 7.1 million USD annually. In addition, it is estimated that paper consumption will be reduced by 2,080 metric tons and greenhouse gas emissions by 1,960 metric tons annually. The presented optimization framework serves as a decision support tool for designing packaging boxes at large e-commerce warehouses.

The facility layout problem (FLP) involves arranging departments on a shop floor to optimise specific objectives, traditionally focussing on pairwise flows between departments. However, these methods often underestimate total travel distances, especially when flows involve multiple input/output points and visits to more than two departments. To address this, connected movements - actual routes taken by transporters - must be considered. This study uses data captured from an Internet of Things (IoT) network and stored on cloud servers to analyze worker movements and accurately calculate travel distances. A mixed-integer programming model is proposed to minimise total travel distance using connected movements as input. Due to the problem's complexity, a biased random key genetic algorithm is employed to find optimal layouts. A case study at a fertiliser production company demonstrates the effectiveness of the approach, achieving a 15% reduction in travel distance compared to layouts generated by traditional methods. The IoT-enabled method also minimises productivity losses by optimising worker movements. While the study focuses on fertiliser manufacturing, the findings are applicable to other settings, such as warehousing, where complex movement sequences and multiple IO points are common in processes like picking, packing, and shipping.

This study investigates the effect of safety-specific transformational leadership (SSTL) on the performance outcomes of safe driving and driving productivity in both long and short-haul truck cargo transport. We conduct our study in the context of a hazardous material (HAZMAT) Indian transport company using a sample of 1,196 trips across 104 unique routes, and driven by 71 truck drivers over a 30-month span. We establish that SSTL is beneficial for truck driving productivity as it positively influences driving productivity in long-haul trips. There is no conclusive evidence of a negative effect on the productivity in short-haul trips. Furthermore, our results show that more experienced drivers are also more likely to indulge in risky driving behavior. Our findings have immediate practical applications for transport companies that wish to promote operational safety, while safeguarding and even improving operational productivity.

In many countries, vehicle replacement policies are implemented to reduce the average age of the vehicles on the road. Through these policies, policymakers typically aim to reduce emissions and to stimulate demand for automobiles through vehicle renewal. Not much is known however, about the more detailed operational consequences of vehicle age in truck transportation. In this study, we empirically address this issue by analyzing data obtained from 27 thousand trips made by 916 drivers in 355 unique trucks, over a period of 346 days. Using this data, we test the relationship between truck age and driver retention, productivity, and unsafe driving behavior. Our results demonstrate that truck age significantly impacts driver turnover, with every additional year of truck age relating to an approximate 5