E-commerce Warehouse Research Articles

EXPRESS: A Machine Learning Approach to Solve the E-commerce Box-Sizing Problem

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 2080 metric tons and greenhouse gas emissions by 1960 metric tons annually. The presented optimization framework serves as a decision support tool for designing packaging boxes at large e-commerce warehouses.

Framework for E-commerce Warehouse Site Selection Evaluation: The Impact of Consumer Repurchase Intention

Supply chains play a decisive role in the development of e-commerce. Although site selection problem for e-commerce warehouses is an important aspect of supply chain management, this topic has received insufficient research attention. This paper fills this research gap by proposing an evaluation framework for the e-commerce warehouse site selection. This framework incorporates consumer repurchase intention (CRI) as a key factor in the evaluation to analyze its impact. This study also employs the matter-element evaluation model to classify the levels of alternatives, and then adopts the probabilistic linguistic terms set to optimize the research results. The feasibility and effectiveness of the proposed evaluation framework is validated through a case study of e-commerce warehouse site selection in China. The results of the case study demonstrate the performance of the proposed framework and highlight the significant impact of CRI on the evaluation outcomes. The proposed framework makes full use of the evaluation information and significantly improves the differences between adjacent alternatives, which can enhance the credibility and accuracy of decision results to provide a more reliable alternative ranking for decision makers.

Picker routing in scattered storage warehouses: an evaluation of solution methods based on TSP transformations

AbstractTo optimize their order fulfillment processes, many e-commerce warehouses employ a storage assignment strategy known as scattered or mixed-shelves storage. Under this approach, unit loads of homogeneous products are divided, and individual pieces are stored in various shelves throughout the warehouse. This arrangement ensures that products that appear together on unpredictable pick lists are stored in close proximity somewhere in the huge warehouses, reducing the travel distance for pickers. Despite these advancements, efficiently guiding pickers through the warehouse remains a significant planning challenge. Since the same products can be found in multiple storage positions, the traditional picker routing problem becomes more complex, as an additional selection task arises regarding which shelf to retrieve each requested product from. While previous research has developed several tailor-made solution algorithms, we demonstrate that known transformation schemes used for different variants of the well-known Traveling Salesman Problem (TSP) can be utilized to convert the single picker routing problem with scattered storage (SPRP-SS) into a classical TSP. This approach enables us to leverage the extensive array of state-of-the-art TSP solvers. The purpose of this paper is to explore the performance of these solvers when applied to solving the SPRP-SS. Through our computational study, we found that existing TSP solvers exhibit good performance, allowing near-optimal solutions to be obtained in less than a second for real-world scale SPRP-SS instances. Moreover, the efficiency of these TSP solvers remains unaffected by the number of cross aisles in the warehouse. Consequently, we exploit this flexibility to investigate the impact of cross aisles on picking performance in scattered storage warehouses.

Synchronizing order picking and delivery in e-commerce warehouses under community logistics

The interplay between inbound order picking and outbound delivery operations is imperative. Nevertheless, these two operations are generally studied in a separate manner, and most existing studies on inbound order picking do not consider the influence of picking operations on downstream delivery operations. This paper addresses the order picking problem in e-commerce warehouses from a new perspective, i.e., towards achieving picking and delivery synchronization. We define a Dynamic Order Picking Problem with Delivery Decisions (DOPP-DD), where picking decisions are dynamically determined, given periodically received community logistics (CL) delivery decisions, to minimize vehicle waiting time at loading docks. The DOPP-DD is modeled as a Markov Decision Process and a selective order picking (SOP) policy is proposed to solve it, leveraging a convolutional neural network (CNN) to predict near-future delivery decisions. Our numerical study showcases the CNN’s high accuracy toward predicting the CL delivery decisions. We compare the performance of the SOP against three benchmark policies across 23 instances. The results reveal that the SOP policy outperforms the others in instances with limited picking capability and demonstrates higher robustness toward limited picking resources, highlighting its potential to alleviate overstock issues and foster a more synchronized workflow in order picking and delivery operations.

A reinforcement learning-based hyper-heuristic for AGV task assignment and route planning in parts-to-picker warehouses

Globally, e-commerce warehouses have begun implementing robotic mobile fulfillment systems (RMFS), which can improve order-picking efficiency by using automated guided vehicles (AGVs) to realize operations from parts to pickers. AGVs depart from their initial points, move to a target rack position, and subsequently transport racks to picking stations. The AGVs return the racks to their original positions after the workers pick them up. When all tasks are completed, the AGVs return to their starting point. In this context, the main challenge is the task assignment and route planning of multiple AGVs to minimize travel times. We formulate a mixed-integer linear programming (MILP) model with valid inequalities to solve small problem instances optimally. We introduce a reinforcement learning (RL)-based hyper-heuristic (HH) framework to solve large instances to near-optimality. A typical HH framework comprises two levels: high-level heuristics (HLH) and low-level heuristics (LLH). The framework starts from an initial solution and improves iteratively through LLHs, while the HLH invokes a selection strategy and an acceptance criterion to generate a new solution. We propose a novel selection strategy based on the improved Multi-Armed Bandits algorithm called Co-SLMAB and Exponential Monte Carlo with counters (EMCQ) as the acceptance criterion. The corresponding collision avoidance rules are then formulated for different conflicts to construct a conflict-free traveling route for AGVs. Besides testing the proposed framework’s effectiveness in real-life warehouse layouts, we perform extensive computational experiments and a thorough sensitivity analysis. The results show that (i) the proposed valid inequalities aid in obtaining better lower bounds and significantly speed up the solution process; (ii) the Co-SLMAB-HH framework is quite competitive compared to CPLEX, outperforming the other tested hyper-heuristics and the problem-specific heuristic regarding convergence and computation time; and (iii) a pool of LLHs consisting of a wide range of different operators is advantageous over a limited set of simple operators while solving problems using hyper-heuristics.

Enhancing E-Commerce Warehouse Management

E-commerce has revolutionized the retail industry, necessitating efficient warehouse management systems to handle the increasing volume of online orders. This paper explores various strategies, technologies, and challenges associated with e-commerce warehouse management. It examines the significance of effective warehouse operations in meeting customer expectations, optimizing inventory, and minimizing costs. Furthermore, it discusses emerging technologies such as automation, robotics, and artificial intelligence (AI) in streamlining warehouse processes. Challenges such as scalability, labor management, and sustainability are also addressed, providing insights into future directions for e-commerce warehouse management.

Enhancing E-Commerce Warehouse Order Fulfillment Through Predictive Order Reservation Using Machine Learning

Enhancing E-Commerce Warehouse Order Fulfillment Through Predictive Order Reservation Using Machine Learning

Simultaneous allocation and sequencing of orders for robotic mobile fulfillment system using reinforcement learning algorithm

Robotic Mobile Fulfillment Systems (RMFS) can benefit large e-commerce warehouse operations significantly. To fulfill the orders received, RMFS deploys mobile robots to carry shelves back and forth from the storage area to the picking station. Order allocation and sequencing for mobile robots is a complex yet critical task as it influences the distance traveled by mobile robots in fulfilling the orders, i.e., appropriately allocating and sequencing the orders. In this paper, a Simultaneous Allocation and Sequencing of Orders Reinforcement Learning (SASORL) algorithm is proposed to minimize the distance traveled by mobile robots. Unlike existing methods, the SASORL algorithm optimizes order allocation and sequencing concurrently, significantly reducing mobile robot travel distance. The proposed SASORL algorithm encompasses three sets, namely state, action, and reward/penalty. The state set comprises the orders fulfilled, whereas the action set contains the orders yet to be fulfilled. The distance traveled by the mobile robot as a result of the orders allocated and sequenced is taken as the penalty for the proposed SASORL algorithm. As the proposed SASORL algorithm simultaneously allocates and sequences the orders to the mobile robots, the action set depletes, the state set enlarges, and the penalty updates until the action set becomes null. Each episode restarts with the learned experience of the prior episodes, and after completing a few episodes, the SASORL algorithm is capable of generating an optimal order allocation and sequence that commits the minimum travel distance to the mobile robots. SASORL algorithm is superior to widely adopted soft computing techniques when orders are randomly distributed. This superiority is evidenced by a 26% reduction in the maximum distance traveled by all mobile robots, a 54% reduction in the standard deviation of the distance traveled by the mobile robots, and a marginal increase of 7% in the total distance traveled by all mobile robots. Moreover, the SASORL algorithm outperforms soft computing techniques by 44% in terms of computation time.

Gaussian process-based storage location assignments with risk assessments for progressive zone picking systems

E-commerce warehouses are under constant pressure to adapt their order picking systems and reassign product storage locations to meet fluctuating customer demands. Most existing approaches optimize storage location reassignments based on customer orders and operational configurations to maintain high order picking performance. This paper presents a Gaussian process surrogate model (GPSM) approach to predict the performance metrics for storage location reassignments. The GPSM estimates the expected flow time of orders from the historical data on previous storage location assignments and aids in identifying the new assignments that yield the minimum estimated average flow times. Management can also take advantage of the GPSM’s uncertainty quantification capability to assess the probability of improvement for a given storage reassignment and its implementation. The proposed model and assignment policy are validated using discrete-event simulations and industrial data. Experimental results demonstrate that the GPSM can improve expected flow time by 7.51% and reduce unnecessary reassignment operations by 43.25%.

Warehouses without aisles: Layout design of a multi-deep rack climbing robotic system

In the last decade, different multi-deep rack climbing robotic (MRR) systems have been introduced, particularly in e-commerce warehouses. These systems have great benefits, as aisles are no longer needed, allowing a high storage density on a small footprint. They only need vertical channels through which battery-powered robots can climb the racks, retrieve totes from any desired position, and bring them to a workstation. This paper studies two novel MRR system layouts: the cross and the compact layout. In addition, we compare performance with the more traditional aisle-based layout. The system performance, particularly operational cost and energy consumption, depends on these system layouts. The paper establishes queuing network models to investigate the trade-off between storage capacity and throughput of the system with these three layouts, taking robot blocking prevention into account. We compare the throughput, storage density, horizontal travel time, and energy consumption of the system. The results show that, in most cases, the compact layout outperforms other layouts on throughput. For energy consumption, the choice of layout depends on the footprint. We formulate a model to assist warehouse managers in choosing a layout of minimum annual operational cost, with a required storage and throughput capacity. We also compare the MRR system with an alternative robotic compact storage and retrieval system on operational cost and energy consumption. The MRR system appears to always have lower energy consumption and operational cost.

Retrieval sequencing in autonomous vehicle storage and retrieval systems

ABSTRACT Autonomous vehicle storage and retrieval systems (AVS/RSs) are widely used in e-commerce warehouses due to their high throughput and flexibility. In such systems, storage and retrieval transactions are performed by lifts and vehicles. This paper focuses on the sequencing retrievals problem in an AVS/RS, which is an important problem for daily operations. We formulate this sequencing problem as a mixed-integer program to determine a retrieval sequence for the lift and the vehicles, one that minimises the makespan. A dynamic programming approach is proposed to solve the sequencing problem to optimality. However, the solution time of the dynamic programming method is exponentially increasing in the number of retrieval requests. To be more practical, we present a beam search heuristic that can solve large-sized instances in reasonable time. Computational experiments verify that near-optimal solutions can be found by the beam search heuristic. Compared to commonly used heuristics and straightforward heuristics, the beam search decreases the makespan by up to 15%. Finally, we analyse how vehicle modes impact the makespan, showing evidence that a small makespan can be achieved when considering a realistic mode of vehicles.

Human-centric investigation of a robotic forward–reserve warehouse

Order picking has long been identified as a time- and cost-intensive task in almost every supply chain. One interesting approach to accelerate the picking process is to separate the warehouse into forward and reserve areas, where the forward area holds a limited amount of inventory for fast picking that is replenished by the stock in the reserve area. To date, the replenishment operation is often conducted manually or using goods-to-person technologies such as automated storage/retrieval systems. Recently, a newly developed market-ready technology called picking robots, which are a kind of autonomous mobile robots, was implemented in e-commerce warehouses. They can be integrated into existing manual order picking systems without modifying the warehouse layout. In this study, we simulate a case wherein picking robots are applied for replenishment tasks to avoid stock shortages in fast-picking areas, thereby supporting the picking activities of humans. The aim is to reduce the human workload during the order picking process. We investigate various scenarios with different human–robot team configurations and demand patterns. The results indicate that such a setting can improve working conditions by simultaneously reducing human energy expenditure and costs per pick. The contribution of this study is to introduce these robots to forward–reserve warehouses, obtain preliminary results regarding system performance, and discuss related opportunities for future work.

Order-Picking Efficiency in E-Commerce Warehouses: A Literature Review

With the vigorous development of e-commerce, efficient order picking in e-commerce warehouses has attracted the attention of many scholars. To analyze the issues about order-picking efficiency currently being studied by relevant scholars in e-commerce warehouses, this paper reviews the literature on the application of order-picking strategy and efficiency optimization direction from 2020 to 2022. That mainly falls into two categories of picking systems: “picker-to-parts” and “parts-to-picker”. In the “picker-to-parts” picking system, more attention is paid to the picking strategies of storage assignment and order batching and the integration of multiple picking strategies. In contrast, in the “parts-to-picker” picking system, the main attention is on the man-machine cooperation in the Mobile Robot Fulfillment System (RMFS) and the Automated Storage and Retrieval System (AS/RS), as well as the coordination of the picking station. Further, this paper proposes future research directions for two categories of picking systems: further studying the order splitting strategy and order delivery issues; considering the dynamic uncertainties; combining the automated picking system with different picking strategies, and so on.

Warehouse Location for Product Distribution by E-Commerce in Brazil: Comparing Symmetrical MCDM Applications

The crisis that affected Brazil as a result of the coronavirus disease 2019 (COVID-19) pandemic impacted several businesses. However, the distribution of products sold by e-commerce is one business that had increased its demand on a large scale. The Southeastern Region of Brazil has a wide territorial extension, a high number of inhabitants, and difficulties with transport logistics. Thus, this work presents a solution for choosing a location for the installation of an e-commerce warehouse, considering three possible locations, in two different Brazilian states: Rio de Janeiro and Sao Paulo. Our main objective is to identify policies that allow a balance in the trade-off between risks and service levels. Specifically, we seek to decrease logistics costs and ensure improvements in services. The choice of a location for a warehouse is a decision problem that can be solved with multiple criteria decision-making (MCDM). Three methods of MCDM have been applied: Fuzzy Decision Approach (FDA), Multi-Attribute Utility Theory (MAUT), and Multi-Attribute Value Theory (MAVT). The reasons for the application of the MAUT and MAVT methods originate from their easy understanding and use, while the concepts of FDA address the consideration of uncertainty. To maintain the simplicity of MCDM application, the FDA model is symmetric, even with symmetrical and asymmetrical fuzzy sets. Important criteria were the distance from capital cities, distance to ports, distance to airports, distance to highways, and the costs of land. Three different locations were considered for the installation of the warehouse. All three methods, FDA, MAUT, and MAVT, resulted in Location 1 as the best location for the installation of the warehouse for the distribution of e-commerce products. A discussion on conceptual qualitative divergences and the small quantitative divergence of results is also included in the conclusions.

Enhance picking viability in E-commerce warehouses under pandemic

The COVID-19 pandemic has caused critical challenges for e-commerce warehouses that strive to fulfill surging customer demand while facing a high virus infection risk. Current literature on picking optimization overlooks warehouse safety under pandemic conditions. Meanwhile, scattered storage and zone-wave-batch picking have been used in parallel by many large e-commerce warehouses, these two operational policies have not been considered together in picking optimization studies. This paper fills these gaps by solving an order batching problem considering scattered storage, zone-wave-batch picking, and pickers’ proximity simultaneously. We formulate and solve the mathematical model of the discussed problem and propose the Aisle-Based Constructive Batching Algorithm (ABCBA) to help warehouses pick more efficiently and safely. Experiments with extensive datasets from a major third-party logistics (3PL) company show that, compared to the current picking strategy, ABCBA can reduce the total picking time and the virus infection risk due to pickers’ proximity by 46% and 72%, respectively. Compared to other heuristics like tabu + nLSA3 (Yang, Zhao, and Guo 2020), ABCBA gets better results using less computation time.

A variable neighborhood search approach to solve the order batching problem with heterogeneous pick devices

In this paper, we discuss the order batching problem in manual order picking systems. In such systems, order pickers move through a warehouse to collect items that are requested by customers. Customer orders have to be grouped into picking orders of limited size to ensure that the total length of the picker tours to collect all items is minimized. Motivated by real-world restrictions, we assume that the pick devices are heterogeneous, i.e., it is not necessarily possible to transport each customer order with each device. Since the order batching problem is NP-hard, we propose both serial and parallel versions of a variable neighborhood search (VNS) scheme to tackle large-sized problem instances in a reasonable amount of time. Moreover, the VNS scheme is hybridized with a set partitioning problem formulation. Based on computational experiments, we show that the proposed VNS scheme is a very fast approach that is able to provide solutions for benchmark instances with homogeneous pick devices that are of comparable quality or even partly superior to the best performing approaches from the literature. New benchmark instances for heterogeneous pick devices are proposed. High-quality solutions are obtained for these instances. We also consider large-scale problem instances from a real-world e-commerce warehouse. The VNS scheme outperforms the heuristic applied for order batching in this warehouse. Moreover, we demonstrate that a warehouse setup with heterogeneous pick devices is highly beneficial with respect to operating cost savings.

Optimal Picking Policies in E-Commerce Warehouses

In e-commerce warehouses, online retailers increase their efficiency by using a mixed-shelves (or scattered storage) concept, where unit loads are purposefully broken down into single items, which are individually stored in multiple locations. Irrespective of the stock keeping units a customer jointly orders, this storage strategy increases the likelihood that somewhere in the warehouse the items of the requested stock keeping units will be in close vicinity, which may significantly reduce an order picker’s unproductive walking time. This paper optimizes picker routing through such mixed-shelves warehouses. Specifically, we introduce a generic exact algorithmic framework that covers a multitude of picking policies, independently of the underlying picking zone layout, and is suitable for real-time applications. This framework embeds a bidirectional layered graph algorithm that provides the best known performance for the simple picking problem with a single depot and no further attributes. We compare three different real-world e-commerce warehouse settings that differ slightly in their application of scattered storage and in their picking policies. Based on these, we derive additional layouts and settings that yield further managerial insights. Our results reveal that the right combination of drop-off points, dynamic batching, the utilization of picking carts, and the picking zone layout can greatly improve the picking performance. In particular, some combinations of policies yield efficiency increases of more than 30% compared with standard policies currently used in practice. This paper was accepted by Chung Piaw Teo, optimization.

A joint optimisation of multi-item order batching and retrieving problem for low-carbon shuttle-based storage and retrieval system

Shuttle-based storage and retrieval system (SBS/RS) is a smart warehouse system evolved from automated storage and retrieval system. It is used for the storage and order picking of products in e-commerce warehouses because of its flexibility and low-carbon feature. This paper investigates the multi-item order batching and retrieving problem for SBS/RS, in which orders containing different items are batched and processed by the system. A mixed-integer linear programming model is proposed to jointly optimise the batching and retrieving assignment, and a tabu search algorithm along with an efficient algorithm for generating retrieving assignments are developed for problem-solving. Numerical experiments are then conducted to validate the efficiency of the proposed algorithm as well as the carbon emission reduction potential of the developed model.

Hybrid modeling of collaborative freight transportation planning using agent-based simulation, auction-based mechanisms, and optimization

The sharing economy is a peer-to-peer economic model characterized by people and organizations sharing resources. With the emergence of such economies, an increasing number of logistics providers seek to collaborate and derive benefit from the resultant economic efficiencies, sustainable operations, and network resilience. This study investigates the potential for collaborative planning enabled through a Physical Internet-enabled logistics system in an urban area that acts as a freight transport hub with several e-commerce warehouses. Our collaborative freight transportation planning approach is realized through a three-layer structured hybrid model that includes agent-based simulation, auction mechanism, and optimization. A multi-agent model simulates a complex transportation network, an auction mechanism facilitates allocating transport services to freight requests, and a simulation–optimization technique is used to analyze strategic transportation planning under different objectives. Furthermore, sensitivity analyses and Pareto efficiency experiments are conducted to draw insights regarding the effect of parameter settings and multi-objectives. The computational results demonstrate the efficacy of our developed model and solution approach, tested on a real urban freight transportation network in a major US city.

Intralogistics synchronization in robotic forward-reserve warehouses for e-commerce last-mile delivery

This paper treats intralogistics processing in a typical robotic forward-reserve e-commerce warehouse involving a reserve area with manual order-picking operations and a forward area with robotic parts-to-picker order sorting operations. It is critical to synchronize intralogistics operations between the two areas considering delivery requirements to optimize the performance in terms of makespan and costs. This challenge is formulated as a delivery-driven intralogistics synchronization (DDIS) problem. This paper develops a tailored variable neighborhood search solution method. A series of comprehensive numerical experiments under various scenarios show the superiority and stability of the DDIS as benchmarked with sequential approaches. A substantial reduction can be achieved in both makespan and forward area size, indicating significantly improved intralogistics operational efficiency and space utilization. Managerial insights are also discussed for specific action plans regarding the market size, the labor force, and the trolley configuration.