Method and system for optimizing a lease of a shared pallet

Abstract

The present application relates to a kind of shared tray lease scheme optimization method and system, belong to logistics equipment sharing technical field, disclose is the lease scheme optimization method and system of shared tray, by obtaining the historical leasing data of each node of supply chain, real-time inventory data and forecast demand data, based on RFID technology collection tray full life cycle track data, establish tray flow digital twin model, mapping whole process state, by improved genetic algorithm to tray scheduling path, inventory distribution and lease pricing are multi-objective optimization, realize the intelligentization, high efficiency and precision of shared tray lease whole process.It solves the problems of inefficient scheduling, rigid pricing, lack of closed-loop management and lack of multi-end collaboration in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of logistics equipment sharing technology, and specifically relates to a method and system for optimizing shared pallet rental schemes. Background Technology

[0002] In supply chain management, pallets, as critical logistics carriers, directly impact logistics costs and management efficiency through their recycling efficiency. Existing shared pallet rental solutions suffer from several shortcomings: First, pallet scheduling relies on manual planning, making dynamic optimization based on real-time inventory, network distribution, and order demand difficult, leading to both resource idleness and supply-demand imbalances. Second, rental pricing uses fixed rules, failing to accurately calculate based on multiple factors such as customer credit rating, rental period, and pallet damage rate, affecting platform revenue and customer satisfaction. Third, pallet lifecycle management lacks a closed loop; data fragmentation across rental, usage, recycling, and maintenance hinders real-time alerts and efficient handling of anomalies. Fourth, multi-terminal collaboration (e.g., platform, network, client, and app) suffers from information delays, resulting in slow order response and poor process integration.

[0003] While RFID-based shared pallet systems have achieved basic information management, existing technologies have not fully explored the value of data and lack optimization mechanisms, making it difficult to meet the needs of efficient supply chain operation. Summary of the Invention

[0004] This invention provides a method and system for optimizing the rental scheme of shared pallets, which solves the problems of inefficient scheduling, rigid pricing, lack of management closed loop, and insufficient multi-terminal collaboration in the prior art. By using an improved genetic algorithm to optimize pallet scheduling path, inventory distribution and rental pricing in multiple objectives, the invention achieves intelligent, efficient and precise sharing of the entire pallet rental process.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: An optimization method for a shared pallet rental scheme includes the following steps: Acquire historical leasing data, real-time inventory data, and forecasted demand data for each node in the supply chain; Based on RFID technology, the entire lifecycle trajectory data of pallets is collected to establish a digital twin model of pallet circulation, which maps the entire process status of pallets from warehousing, leasing, transportation, recycling to repair / scrapping; An improved genetic algorithm is used to perform multi-objective optimization of pallet scheduling paths, inventory distribution, and rental pricing.

[0006] Furthermore, in the step of acquiring data from each node of the supply chain, edge computing nodes are deployed at warehouses / outlets to collect pallet entry and exit data, equipment operating status data, and environmental temperature and humidity data in real time. The pre-processed data is uploaded to the cloud platform via a 5G communication module. The pre-processing process includes data deduplication, outlier removal, and format standardization. Outlier removal is based on the 3σ principle to identify loss rates and inventory quantity data that exceed the normal range.

[0007] Furthermore, in the step of establishing a digital twin model of pallet circulation, the unique pallet identifier collected by RFID is bound to the virtual pallet in the digital twin model, and the physical status parameters of the pallet are synchronized in real time. Physical status parameters include wear level, load history, location coordinates and associated order information. Time series analysis is used to predict the remaining service life of the pallet. When the remaining service life is lower than the first preset threshold, maintenance reminders or scrap warnings are automatically triggered.

[0008] Furthermore, in the improved genetic algorithm optimization step, the encoding method adopts real number encoding, and the pallet scheduling path, inventory allocation ratio and pricing coefficient are used as gene segments. The fitness function is constructed as follows: transportation cost weight coefficient × transportation cost value + inventory turnover rate weight coefficient × (1 - inventory turnover cycle) + customer satisfaction weight coefficient × customer satisfaction value. The weight coefficients are determined by the analytic hierarchy process. In genetic operations, adaptive crossover and mutation probabilities are introduced and dynamically adjusted according to the number of generations of population evolution to avoid the algorithm getting trapped in local optima.

[0009] Furthermore, this includes a dynamic adjustment process for lease pricing: a pricing factor system is constructed based on customer credit rating, lease volume, lease period, and historical cooperation frequency. Customer credit rating is comprehensively assessed through account performance records, settlement timeliness, and pallet loss rate. A gradient boosting regression model is used to train the pricing factors and generate personalized pricing schemes. When the customer's rental volume exceeds the second preset threshold or the rental period covers the industry's off-season, the discount coefficient is automatically adjusted.

[0010] Furthermore, this includes an intelligent order matching step: based on the pallet type, quantity, delivery address, and usage period requirements in customer orders, combined with optimized inventory distribution data, the system automatically matches the optimal outbound warehouse / outlet and carrier; During the matching process, priority is given to warehouses / outlets that are closest to the customer and have pre-set pallet availability, while also taking into account the carrier's historical on-time delivery rate and transportation cost quotes.

[0011] Further steps include recycling optimization: based on recycling application data initiated by members, geographical distribution of collection points, and pallet status data, the optimal recycling route is planned, and clustering algorithms are used to merge adjacent recycling requests to reduce the empty mileage of recycling vehicles. The recycling pallets are graded and tested, and then allocated to the reuse, repair or disposal channels according to the test results. The test indicators include appearance integrity, structural strength and RFID signal stability.

[0012] Furthermore, this includes intelligent processing steps for abnormal orders: real-time monitoring of rental order status through machine learning models to identify abnormal situations such as order delays, pallet damage, and settlement anomalies; The machine learning model uses the random forest algorithm and is trained and generated based on historical abnormal order data; Corresponding handling strategies are implemented for different types of abnormal triggers. When an order is delayed, the subsequent scheduling plan is automatically adjusted. When a pallet is damaged, the claims process is initiated and a replacement pallet is provided. When a settlement is abnormal, a reminder message is pushed to the finance module.

[0013] Furthermore, this includes optimizing the effectiveness evaluation process: establishing an evaluation indicator system, which includes cost reduction rate, inventory turnover improvement rate, order response time reduction rate, customer complaint rate, and pallet recycling rate; Regularly compare and analyze the optimized actual operating data with the baseline data before optimization, and dynamically adjust the parameters of the genetic algorithm and the optimization target weights based on the evaluation results to achieve continuous iterative optimization of the leasing scheme.

[0014] A shared pallet rental scheme optimization system for performing the above method includes: The data acquisition and preprocessing module is used to acquire historical leasing data, real-time inventory data, and forecasted demand data from each node of the supply chain. The digital twin modeling module, based on the full lifecycle trajectory data of the pallet collected by RFID technology, establishes a digital twin model of pallet circulation, binds the unique pallet identifier collected by RFID to the virtual pallet in the digital twin model, and synchronizes the physical status parameters of the pallet in real time and maps the status of the entire process. The multi-objective optimization module uses an improved genetic algorithm to perform multi-objective optimization of pallet scheduling paths, inventory distribution, and rental pricing; The system also includes a function execution module, comprising a rental pricing dynamic adjustment unit, an order intelligent matching unit, a recycling optimization unit, an abnormal order intelligent processing unit, and an optimization effect evaluation unit; the function execution module is configured to execute the steps of the above methods.

[0015] As can be seen from the above, this invention tracks the status of pallets in real time by establishing a digital twin model of pallet circulation and combining it with an improved genetic algorithm to achieve multi-objective dynamic optimization. This solves the problems of low resource scheduling efficiency, rigid pricing mechanism and lack of full-process management in the existing technology, and has significant advantages in improving supply chain collaboration efficiency and reducing operating costs. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the workflow of the present invention; Figure 2 This is a schematic diagram of the internal structure of the system of the present invention; Figure 3 This is a schematic diagram of the internal structure of the functional execution module of the present invention.

[0019] The following are the diagram labels: 10. Data Acquisition and Preprocessing Module; 20. Digital Twin Modeling Module; 30. Multi-Objective Optimization Module; 40. Function Execution Module; 401. Dynamic Adjustment Unit for Rental Pricing; 402. Intelligent Order Matching Unit; 403. Recycling Optimization Unit; 404. Intelligent Handling Unit for Abnormal Orders; 405. Optimization Effect Evaluation Unit.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application. To better understand the technical solutions of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0022] Shared pallet leasing solutions in supply chain management generally suffer from low resource scheduling efficiency, inflexible pricing mechanisms, lack of closed-loop lifecycle management, and delays in multi-terminal collaborative information. Traditional methods rely on manual planning of scheduling routes, making it difficult to dynamically adjust based on real-time inventory and order demand, resulting in both idle and scarce pallets. Leasing pricing uses fixed rules and fails to dynamically adjust based on customer credit ratings and pallet damage rates, impacting operational profitability. Data at each stage of pallet circulation is isolated, making full lifecycle status tracking impossible and hindering timely handling of anomalies.

[0023] To address the aforementioned issues, existing technologies suffer from limited data acquisition dimensions and lack real-time monitoring of pallet physical status, leading to delayed scheduling decisions. Analysis reveals that establishing a complete data chain is crucial for achieving dynamic optimization. To solve the challenge of multi-objective collaborative optimization, an attempt was made to introduce genetic algorithms into the scheduling model; however, traditional algorithms are prone to getting trapped in local optima. Multiple experiments verified that using real-number encoding combined with adaptive parameter adjustment effectively improves algorithm convergence. To achieve closed-loop management, a proposal was made to bind the physical parameters collected by RFID to a digital twin model, synchronously updating the pallet status.

[0024] Therefore, as Figure 1 As shown in the figure, this embodiment proposes an optimization method for a shared pallet rental scheme, including the following steps: Step S100: Obtain historical leasing data, real-time inventory data, and forecasted demand data for each node in the supply chain; Step S200: Collect pallet trajectory data throughout its entire lifecycle based on RFID technology, establish a digital twin model of pallet circulation, and map the entire process status of the pallet from warehousing, leasing, transportation, recycling to repair / scrapping; Step S300: Perform multi-objective optimization of pallet scheduling path, inventory distribution and rental pricing using an improved genetic algorithm.

[0025] Historical rental data refers to pallet rental records at each node within past periods, obtained through database query interfaces, used to analyze customer rental behavior patterns. Real-time inventory data refers to the number of pallets currently available in the warehouse or network, collected in real time by IoT sensors, providing a basis for dynamic scheduling. Forecasted demand data refers to the pallet demand in various regions within a future time period, generated using time series forecasting models to predict resource gaps. Full lifecycle trajectory data refers to records of pallet status changes during circulation, collected at key nodes using RFID readers to capture location, timestamps, and operation types, forming a complete circulation chain. The digital twin model is a dynamic pallet mapping model constructed in virtual space, using 3D modeling technology to achieve real-time synchronization of physical parameters with the virtual model, supporting full-process status visualization. The improved genetic algorithm is an optimized swarm intelligence algorithm, using real-number encoding to express scheduling paths and inventory allocation schemes, and improving global search capabilities through adaptive crossover and mutation probabilities.

[0026] Specifically, historical data from each node in the supply chain is transmitted to the central processing platform via data interfaces, forming a multidimensional dataset. RFID readers automatically scan electronic tags when pallets enter or leave the warehouse, recording time, location, and operation type to generate full lifecycle trajectory data. The digital twin model receives real-time data and updates the virtual pallet's status parameters; for example, when a pallet is transported to a transfer station, the model synchronously updates its location coordinates and transportation time. The improved genetic algorithm encodes the scheduling path as a sequence of real numbers, with each gene segment corresponding to the selection order of warehouse nodes, and the inventory allocation ratio encoded as the resource proportion of each node. During algorithm iteration, the fitness function calculates the weighted sum of transportation costs, inventory turnover rate, and customer satisfaction, and selects high-quality individuals using a roulette wheel strategy. Crossover operations use a two-point crossover method to exchange parent gene segments, and mutation operations randomly adjust some gene values ​​to increase population diversity.

[0027] Compared to existing technologies, traditional solutions rely on human experience to formulate scheduling plans, making it difficult to cope with dynamically changing demand fluctuations. This solution, however, utilizes genetic algorithms to achieve multi-objective automatic optimization, significantly improving decision-making efficiency. In existing technologies, pallet status data is scattered across independent systems, failing to form a closed-loop management process. This solution uses a digital twin model to achieve real-time mapping between the physical world and virtual space, enabling timely identification of abnormal states. Traditional pricing models use fixed coefficients to calculate rental fees; this solution combines real-time inventory data and forecasted demand to dynamically adjust pricing strategies, enhancing market adaptability.

[0028] Through the aforementioned technical means, this embodiment can dynamically optimize pallet scheduling routes, reducing the empty load rate of transport vehicles; achieve reasonable allocation of inventory resources, reducing the risk of regional shortages; establish a full lifecycle data chain, improving the accuracy of pallet status tracking; and balance cost control and customer satisfaction through a multi-objective collaborative optimization mechanism. The leasing pricing strategy can be flexibly adjusted according to market demand, enhancing the platform's profitability. The digital twin model provides data support for anomaly early warning, shortening fault response time.

[0029] Furthermore, this embodiment proposes to use edge computing nodes deployed at warehouses or outlets in the data collection steps of each node in the supply chain to collect pallet entry and exit data, equipment operation status data and environmental temperature and humidity data in real time. The pre-processed data is uploaded to the cloud platform through a 5G communication module. The pre-processing process includes data deduplication, outlier removal and format standardization. Outlier removal is based on the 3σ principle to identify loss rate and inventory quantity data that exceed the normal range.

[0030] Edge computing nodes refer to distributed computing devices deployed in physical warehouses or service outlets. Specifically, they can be implemented using embedded industrial control computers combined with IoT gateways to process local sensor data locally, reducing transmission latency. 5G communication modules refer to wireless transmission devices supporting fifth-generation mobile communication technology. Specifically, they can be implemented using industrial-grade communication modules integrating 5G basebands to ensure low-latency transmission of high-concurrency data. Data deduplication refers to the process of eliminating duplicate collections or redundant records. Specifically, it can be implemented using hash value comparison algorithms to reduce the storage resources occupied by invalid data. Outlier removal refers to filtering out abnormal data that exceeds a reasonable range. Specifically, it can be implemented using the 3σ principle based on normal distribution to calculate data fluctuation ranges and automatically filter out outlier records that deviate from the mean by more than three standard deviations. Format standardization refers to unifying the storage structure of multi-source heterogeneous data. Specifically, it can be implemented using JSON format conversion tools to convert raw data collected by different devices into standardized data packets with unified fields.

[0031] Specifically, edge computing nodes directly connect to RFID readers, temperature and humidity sensors, and the warehouse management system within the warehouse, acquiring real-time data such as pallet entry / exit timestamps, forklift operating status, and warehouse environmental parameters with millisecond-level response times. During the data collection process, the data deduplication module eliminates redundant information caused by repeated sensor reports by comparing the hash values ​​of adjacent data records in the time series. The outlier removal module performs statistical distribution analysis on data showing sudden changes in inventory quantity or abnormally high pallet loss rates, dynamically calculating the normal value range using the 3σ principle and automatically filtering out abnormal data points exceeding the threshold range. After standardization, the data packets are transmitted to the cloud via a 5G communication module, with transmission latency controlled within 50 milliseconds, ensuring real-time synchronization of dynamic supply chain data.

[0032] Compared to existing technologies, traditional supply chain data collection relies on a central server polling data from each node, resulting in minute-level latency and susceptibility to network fluctuations. This solution utilizes edge computing nodes for on-site data processing, reducing collection response time from minutes to milliseconds, while leveraging 5G communication to ensure stable transmission of massive amounts of data. Existing technologies often employ manual sampling to handle outlier data; this solution uses a 3σ algorithm for automatic identification and filtering of outliers, improving data cleaning efficiency by two orders of magnitude and effectively preventing erroneous data from interfering with subsequent optimization algorithms.

[0033] This invention addresses the core issues of insufficient real-time data collection and low data quality at supply chain nodes through this technological approach. The distributed deployment of edge computing nodes reduces data collection latency to a level where business systems can respond in real time. The 5G communication module ensures efficient transmission of preprocessed data, and the statistical outlier removal mechanism significantly improves the accuracy of inventory and loss data. Standardized format processing eliminates structural barriers to multi-source data fusion. These technologies collectively construct a high-quality data collection system, providing reliable data support for pallet scheduling path optimization and dynamic inventory allocation.

[0034] Furthermore, this embodiment proposes that in the step of establishing a digital twin model of pallet circulation, the unique pallet identifier collected by RFID is bound to the virtual pallet in the digital twin model, and the physical status parameters of the pallet are synchronized in real time. The physical status parameters include wear level, load history, location coordinates and associated order information. The remaining service life of the pallet is predicted through time series analysis, and maintenance reminders or scrap warnings are automatically triggered when the remaining service life is lower than a preset threshold.

[0035] Among these, RFID unique identification refers to the identification code assigned to each physical pallet through radio frequency identification technology. Specifically, this can be implemented using passive electronic tags conforming to the ISO 18000-6C standard, used for automatic identification and tracking of individual pallets in the supply chain. Digital twin models refer to virtualized pallet objects constructed through the fusion of 3D modeling and IoT data. Specifically, this can be implemented using the Unity3D engine and the MQTT protocol, used to map real-time changes in the physical pallet's state. Wear degree, among the physical state parameters, refers to the fatigue damage values ​​of the pallet's surface and structural components. Specifically, this can be calculated using finite element analysis after collecting deformation data with a laser scanner. Load history refers to the load records the pallet has borne in each rental task, which can be obtained by linking pressure sensors with an order database. Time series analysis is a method for dynamically predicting the trends of pallet performance parameters. Specifically, this can be implemented using an ARIMA model combined with a sliding window mechanism to capture the performance degradation patterns of the pallet.

[0036] Specifically, when a pallet enters the warehouse, an RFID reader automatically reads its unique identifier and generates a corresponding virtual pallet instance in the digital twin system. The virtual model receives real-time data from pallet sensors via an IoT gateway, including vibration frequency, load weight, and ambient temperature and humidity parameters for each transport mission. Historical load data is stored as a timestamp sequence and linked to the cargo type and transport mileage information in the order system. Wear and tear is calculated by periodically scanning the pallet structure and comparing it with the initial model, forming a time series of material fatigue coefficients. These parameters are input into a trained ARIMA prediction model, which analyzes the autocorrelation and differential stationarity of historical data to generate a prediction curve for the remaining service life. When the predicted value is lower than a preset maintenance threshold, the system automatically generates a work order and assigns it to the nearest maintenance station; if the predicted value is lower than a scrap threshold, a replacement process is triggered and inventory data is updated.

[0037] Compared to existing technologies, traditional pallet management relies on manual inspections and scheduled maintenance, failing to detect pallet status changes in real time, leading to delayed maintenance responses or over-maintenance. While existing technologies employ RFID for pallet tracking, they lack a virtual-physical twin model, and physical parameter acquisition is limited to location information, lacking dynamic monitoring of structural performance. This solution integrates multi-source sensor data and business system information to construct a comprehensive status assessment system. By using time series analysis to replace fixed-cycle detection mechanisms, it achieves accurate prediction and tiered processing of maintenance needs.

[0038] Through the aforementioned technical means, this invention achieves real-time synchronization between the physical status of pallets and their digital models, solving the data lag problem in traditional management. Time series analysis based on dynamic parameters accurately captures pallet performance degradation trends, avoiding misjudgments caused by fixed thresholds. The dual triggering mechanism of maintenance reminders and scrap warnings allows for differentiated handling based on actual wear and tear, extending the service life of usable pallets while preventing safety hazards. The historical records associated with order information provide a data foundation for analyzing pallet usage patterns, facilitating the optimization of subsequent scheduling strategies and inventory distribution.

[0039] Furthermore, this embodiment proposes that in the improved genetic algorithm optimization step, the encoding method adopts real number encoding, and the pallet scheduling path, inventory allocation ratio and pricing coefficient are used as gene segments; the fitness function is constructed as transportation cost weight coefficient × transportation cost value + inventory turnover rate weight coefficient × (1 - inventory turnover cycle) + customer satisfaction weight coefficient × customer satisfaction value, and the weight coefficients are determined by the analytic hierarchy process; adaptive crossover probability and mutation probability are introduced in the genetic operation, which are dynamically adjusted according to the number of generations of population evolution to avoid the algorithm getting trapped in local optima.

[0040] Real-number encoding refers to directly using the coordinate sequence of the scheduling path, the inventory allocation ratio, and the pricing coefficient as chromosome genes. This can be implemented using continuous variable encoding, ensuring a continuous distribution of multidimensional decision variables in the solution space. The composite fitness function linearly weights the three optimization objectives: transportation cost, inventory turnover rate, and customer satisfaction. The weight ratio of each objective can be determined using the analytic hierarchy process (AHP), for example, by having domain experts compare the importance of objectives pairwise and then calculating the weight coefficients. Adaptive genetic operations refer to the dynamic change of crossover and mutation probabilities with increasing generations. This can be achieved by using an exponential decay function to adjust the crossover probability and dynamically adjusting the mutation probability based on population diversity indicators.

[0041] Specifically, in the initialization phase of the genetic algorithm, the latitude and longitude coordinate sequence of the scheduling path, the inventory allocation percentage of each warehouse, and the pricing coefficient values ​​of different customer groups are encoded as real-valued gene fragments. During the iteration process, the transportation cost is calculated using path distance and unit freight cost, the inventory turnover cycle is determined by the ratio of inventory level to average daily consumption, and the customer satisfaction value is a comprehensive score based on historical order fulfillment rate and complaint rate. The analytic hierarchy process (AHP) assigns weights to the three optimization objectives according to business needs; for example, the weight of transportation cost is set to 0.5, the weight of inventory turnover rate is set to 0.3, and the weight of customer satisfaction is set to 0.2. As the number of generations increases, the crossover probability gradually decreases from the initial value of 0.8 to 0.4 according to an exponential function, while the mutation probability gradually increases from 0.01 to 0.1 based on the population individual similarity index, thereby balancing global search and local optimization capabilities.

[0042] Compared to existing technologies, traditional genetic algorithms, which use binary encoding, struggle to represent continuous variables such as inventory allocation ratios. This solution, however, directly maps the decision variable space using real-number encoding. Existing technologies often employ fixed-weight fitness functions, failing to dynamically adjust target weights based on business priorities. This solution, however, achieves dynamic weight configuration through the analytic hierarchy process (AHP). Conventional genetic algorithms, with their fixed crossover and mutation probabilities, are prone to premature convergence. This method, through an adaptive mechanism, maintains population diversity in the early stages of evolution and enhances local search accuracy in later stages.

[0043] Through the aforementioned technical means, this invention solves the problem of limited solution space caused by imprecise variable expression in multi-objective optimization, and achieves coordinated optimization of scheduling paths, inventory distribution, and pricing strategies. A dynamic weight allocation mechanism balances the conflict between cost control and service quality, avoiding business imbalances caused by single-objective optimization. Adaptive genetic operations effectively prevent the algorithm from getting trapped in local optima, quickly converging to the global optimum under complex constraints.

[0044] Furthermore, a dynamic adjustment procedure for rental pricing is proposed. A pricing factor system is constructed based on customer credit rating, rental volume, rental period, and historical cooperation frequency. Customer credit rating is comprehensively evaluated through account performance records, settlement timeliness, and pallet loss rate. A gradient boosting regression model is used to train the pricing factors to generate personalized pricing schemes. When the customer's rental volume exceeds the preset threshold or the rental period covers the industry's off-season, the discount coefficient is automatically adjusted.

[0045] The pricing factor system refers to a set of variables that affect the rental price, including customer credit rating, rental volume, rental period, and historical cooperation frequency. Specifically, data standardization methods can be used to convert each factor into a value with a unified dimension, and a weighted combination can be used to form the basic parameters for pricing. This system is used to solve the problem that a single factor in traditional pricing cannot reflect the comprehensive value of the customer.

[0046] Customer credit rating refers to a credit score calculated based on account performance records, settlement timeliness, and pallet loss rate. Specifically, the weight of each indicator can be determined using the analytic hierarchy process (AHP), for example: performance records account for 40%, settlement timeliness for 30%, and loss rate for 30%. A weighted summation is then used to generate a credit score from 0 to 100, used to differentiate the credit risk levels of different customers. The gradient boosting regression model is an ensemble learning algorithm that uses pricing factors and transaction prices from historical leasing data as a training set. It iterative optimization of the residuals fits the nonlinear relationship between leasing prices, capturing the complex correlation between batch size, period, and price.

[0047] The preset threshold refers to the critical number of rentals that triggers the bulk discount. Specifically, it can be set to 1.5-2 times the industry average rental volume based on the historical order distribution. For example, when the single rental volume exceeds 200 pallets, the discount coefficient is activated to incentivize large customers to increase their rental scale.

[0048] Among them, the industry off-season refers to the low demand period defined by historical demand fluctuations. Specifically, time series analysis can be used to identify March to May each year as the logistics off-season, which can be used to increase pallet turnover rate through discounts during the low demand period.

[0049] Specifically, customer credit ratings are quantified using order completion rates from account performance records, median overdue days for settlement timeliness, and the historical average of pallet loss rates, forming differentiated credit rating labels. Rental volume, duration, and historical cooperation frequency are normalized to values ​​in the 0-1 range and input into a gradient boosting regression model along with the credit rating. The model generates personalized quotes including a base price and a floating range by fitting the non-linear relationship between historical transaction prices and various factors. When a customer's rental volume exceeds a preset threshold, the system automatically calls the discount coefficient calculation module; for example, a 5% discount is triggered when the rental volume reaches 300 units. If the rental period covers the industry's off-season, an off-season discount coefficient is applied; for example, renting 6 months of service in March can receive an additional 3% discount. This achieves dynamic matching of pricing strategies with customer behavior and market cycles.

[0050] Compared to existing technologies, current shared pallet rental pricing typically employs fixed rates or simple segmented pricing, such as charging a uniform daily rental fee based on pallet type, without considering customer credit differences and rental behavior characteristics. This solution constructs a multi-dimensional pricing factor system, incorporating variables such as credit scores and rental scale into model training. This allows high-credit customers to receive lower base rates, long-term cooperative customers to enjoy cumulative discounts over a period, and customers with excessive loss rates to be subject to a premium coefficient. Furthermore, compared to linear regression methods, the gradient boosting regression model more accurately reflects the non-linear interactions between price and multiple factors, such as the significant changes in price elasticity when the rental volume is between 200 and 300 units.

[0051] By employing the aforementioned technical means, this application addresses the revenue loss problem caused by the crude customer segmentation in traditional pricing mechanisms. Through quantitative assessment of credit rating and leasing behavior, high-risk customers bear higher costs, while high-quality customers receive price incentives, thereby balancing platform revenue and customer satisfaction. The automatic discount mechanism for off-season periods effectively improves the utilization rate of idle pallets and avoids resource waste. Furthermore, the application of a gradient-upgrading regression model allows the pricing strategy to be continuously optimized based on historical data, adapting to changes in market supply and demand.

[0052] Furthermore, an intelligent order matching step is proposed. Based on the pallet type, quantity, delivery address, and usage period requirements in customer orders, combined with optimized inventory distribution data, the optimal outbound warehouse / outlet and carrier are automatically matched. During the matching process, the warehouse / outlet closest to the customer and with a pallet availability rate of ≥90% are prioritized, while also taking into account the carrier's historical on-time delivery rate and transportation cost quotes.

[0053] Among these, a pallet availability rate of ≥90% refers to the proportion of available pallets in the warehouse / outlet relative to the total inventory. This is specifically calculated by collecting pallet availability data from a real-time inventory monitoring system to ensure that the actual inventory of the warehouse / outlet meets order requirements during matching. Historical on-time delivery rate refers to the proportion of orders delivered on time within a preset time period by the carrier. This is specifically calculated by extracting the difference between the carrier's historical order receipt time and the promised time from a logistics data platform to assess the carrier's reliability in transportation timeliness. Transportation cost quotation refers to the fee standard offered by the carrier for a specific transportation distance and cargo volume. This can be obtained through a logistics bidding platform or carrier quotation system to quantify the economics of transportation services.

[0054] Specifically, when a customer submits an order containing pallet type, quantity, delivery address, and usage period, the system uses a geographic information system to analyze the latitude and longitude coordinates of the delivery address and calculates the distance between the customer's location and various warehouses / outlets. It prioritizes selecting the three warehouses / outlets closest to the customer and verifies their current inventory data to ensure pallet availability reaches 90%. If the availability meets the standard, the warehouse / outlet is added to the candidate list. Subsequently, the system retrieves carrier information associated with the candidate warehouses / outlets from the carrier database, extracting their historical on-time delivery rates and current transportation cost quotes. A weighted scoring model is used to comprehensively evaluate the timeliness and economy of the candidate carriers, generating the optimal warehouse / outlet and carrier matching combination, and pushing the matching results to the scheduling system for execution.

[0055] Compared to existing technologies, traditional order matching relies on human experience to select warehouses and carriers, neglecting real-time inventory status and carrier service quality, which can easily lead to matching results deviating from the optimal solution. This method, however, constructs a multi-dimensional matching model by quantitatively evaluating pallet availability, transportation timeliness, and cost parameters, achieving dynamic and accurate matching between order demand and resource supply, effectively eliminating subjective errors in human decision-making.

[0056] Through the above technical solution, this application solves the problem of order response delays caused by the low efficiency of manual matching in the shared pallet rental process. It also optimizes transportation routes and carrier selection, avoiding additional scheduling costs due to insufficient inventory or unstable carrier services. The automated matching mechanism ensures that customers obtain pallet resources that meet their needs in the shortest possible time, reducing overall logistics operating costs.

[0057] Furthermore, the paper proposes to plan the optimal recycling route based on recycling application data initiated by members, the geographical distribution of recycling points, and pallet status data. It also proposes to use a clustering algorithm to merge adjacent recycling requests to reduce the empty mileage of recycling vehicles. At the same time, the paper proposes to conduct graded inspections on recycling pallets and allocate them to reuse, repair, or scrapping channels according to the inspection results. The inspection indicators include appearance integrity, structural strength, and RFID signal stability.

[0058] Clustering algorithms group spatially similar recycling demand points, typically using K-means or DBSCAN algorithms. By calculating the geographical distances between recycling points, task sets are formed, reducing duplicate vehicle routes. Appearance integrity inspection uses image recognition technology to quantitatively assess physical damage such as scratches and deformations on the pallet surface, typically using a convolutional neural network model to determine if the pallet's appearance meets reuse standards. Structural strength testing measures the pallet's load-bearing capacity attenuation using pressure sensors, typically using a dynamic load testing device, to assess whether the pallet's mechanical properties meet safety thresholds. RFID signal stability testing verifies the response frequency and data integrity of electronic tags using a reader, typically using a signal strength threshold judgment method to ensure pallet traceability in subsequent circulation.

[0059] Specifically, the recycling application data includes the location, quantity, and status information of the pallets to be recycled, which, combined with network distribution data, generates an initial recycling task set. A clustering algorithm groups adjacent recycling points into the same task group based on geographical coordinates, forming an optimized multi-path recycling scheme. Recycling vehicles execute recycling operations according to the merged task groups, avoiding empty runs caused by traveling between scattered locations. Recycled pallets enter the inspection phase: appearance integrity inspection identifies the degree of surface damage, structural strength inspection assesses load-bearing capacity degradation, and RFID signal stability inspection determines chip readability. When all three indicators meet the standards, the pallet is assigned to the reuse channel; if only the structural strength fails to meet the standard, it is assigned to the maintenance channel; and if any one indicator significantly exceeds the standard, it is assigned to the scrap channel.

[0060] Compared to existing technologies, traditional recycling route planning relies on manual experience to divide areas, which can easily lead to path intersections or incomplete coverage. Clustering algorithms, however, automatically generate task groups through spatial data analysis, improving route planning efficiency. Existing technologies typically only perform visual inspections of recycling pallets, lacking quantitative evaluation standards. This solution, however, establishes a hierarchical evaluation system through multi-dimensional detection indicators, improving the accuracy of sorting and processing.

[0061] Through the aforementioned technical means, this application can effectively reduce the empty mileage of recycling vehicles, reduce fuel consumption and carbon emissions, and improve pallet sorting and processing efficiency through automated detection processes, avoiding resource waste caused by human error. The tiered detection mechanism ensures that pallets in different states enter the appropriate processing channel, extending the service life of reusable pallets, reducing maintenance costs, and ensuring the timely removal of obsolete pallets.

[0062] Furthermore, the paper proposes to plan the optimal recycling route based on recycling application data initiated by members, the geographical distribution of recycling points, and pallet status data. It also proposes to use a clustering algorithm to merge adjacent recycling requests to reduce the empty mileage of recycling vehicles. At the same time, the paper proposes to conduct graded inspections on recycling pallets and allocate them to reuse, repair, or scrapping channels according to the inspection results. The inspection indicators include appearance integrity, structural strength, and RFID signal stability.

[0063] Clustering algorithms group spatially similar recycling demand points, typically using K-means or DBSCAN algorithms. By calculating the geographical distances between recycling points, task sets are formed, reducing duplicate vehicle routes. Appearance integrity inspection uses image recognition technology to quantitatively assess physical damage such as scratches and deformations on the pallet surface, typically using a convolutional neural network model to determine if the pallet's appearance meets reuse standards. Structural strength testing measures the pallet's load-bearing capacity attenuation using pressure sensors, typically using a dynamic load testing device, to assess whether the pallet's mechanical properties meet safety thresholds. RFID signal stability testing verifies the response frequency and data integrity of electronic tags using a reader, typically using a signal strength threshold judgment method to ensure pallet traceability in subsequent circulation.

[0064] Specifically, the recycling application data includes the location, quantity, and status information of the pallets to be recycled, which, combined with network distribution data, generates an initial recycling task set. A clustering algorithm groups adjacent recycling points into the same task group based on geographical coordinates, forming an optimized multi-path recycling scheme. Recycling vehicles execute recycling operations according to the merged task groups, avoiding empty runs caused by traveling between scattered locations. Recycled pallets enter the inspection phase: appearance integrity inspection identifies the degree of surface damage, structural strength inspection assesses load-bearing capacity degradation, and RFID signal stability inspection determines chip readability. When all three indicators meet the standards, the pallet is assigned to the reuse channel; if only the structural strength fails to meet the standard, it is assigned to the maintenance channel; and if any one indicator significantly exceeds the standard, it is assigned to the scrap channel.

[0065] Compared to existing technologies, traditional recycling route planning relies on manual experience to divide areas, which can easily lead to path intersections or incomplete coverage. Clustering algorithms, however, automatically generate task groups through spatial data analysis, improving route planning efficiency. Existing technologies typically only perform visual inspections of recycling pallets, lacking quantitative evaluation standards. This solution, however, establishes a hierarchical evaluation system through multi-dimensional detection indicators, improving the accuracy of sorting and processing.

[0066] Through the aforementioned technical means, this application can effectively reduce the empty mileage of recycling vehicles, reduce fuel consumption and carbon emissions, and improve pallet sorting and processing efficiency through automated detection processes, avoiding resource waste caused by human error. The tiered detection mechanism ensures that pallets in different states enter the appropriate processing channel, extending the service life of reusable pallets, reducing maintenance costs, and ensuring the timely removal of obsolete pallets.

[0067] Further, optimization effect evaluation steps were proposed, and an evaluation index system was established, including cost reduction rate, inventory turnover improvement rate, order response time reduction rate, customer complaint rate, and pallet recycling rate. The optimized actual operating data was compared and analyzed with the baseline data before optimization on a regular basis. The parameters of the genetic algorithm and the optimization target weights were dynamically adjusted according to the evaluation results to achieve continuous iterative optimization of the leasing scheme.

[0068] The evaluation index system refers to a multi-dimensional set of indicators used to quantitatively evaluate the optimization effect. Specifically, it can use cost reduction rate to reflect changes in transportation costs, inventory turnover improvement rate to measure inventory efficiency, order response time reduction rate to assess service speed, customer complaint rate to reflect customer satisfaction, and pallet recycling rate to characterize resource utilization efficiency. These indicators are collected and calculated through a data statistics module to provide a basis for subsequent parameter adjustments.

[0069] Dynamically adjusting the parameters and optimization target weights of the genetic algorithm refers to modifying the crossover probability, mutation probability, and weight coefficients of each optimization target based on the evaluation results. Specifically, an adaptive algorithm can be used to automatically adjust parameters according to preset rules. For example, when the inventory turnover improvement rate does not meet expectations, the weight coefficient of the inventory allocation strategy can be increased. This adjustment mechanism can avoid the local optima problem caused by fixed parameters.

[0070] Specifically, the optimization effect evaluation step forms a closed-loop feedback mechanism by constructing multi-dimensional indicators. The evaluation indicator system transforms operational data from different dimensions into quantifiable evaluation standards. For example, by comparing order response times before and after optimization, the effectiveness of scheduling path optimization can be determined. The differences between actual operating data and baseline data are analyzed regularly. For instance, an increase in customer complaint rates indicates deficiencies in the current pricing strategy or service process. Based on the results of the difference analysis, the parameters of the genetic algorithm are adjusted, reducing the weight of transportation costs to prioritize customer satisfaction, or increasing the mutation probability to enhance the algorithm's search capabilities. This dynamic adjustment allows the optimization process to respond to business changes in real time. For example, during peak sales seasons, the optimization priority is increased for inventory turnover, while during peak maintenance periods, the focus is on improving pallet recycling rates.

[0071] In a specific implementation, the benchmark data for the evaluation index system can be set as the historical average of the three months prior to optimization, while actual operational data is collected in real time via IoT devices. Comparative analysis can be performed using difference rate calculations, for example, cost reduction rate = (benchmark transportation cost - actual transportation cost) / benchmark transportation cost × 100%. Parameter adjustment strategies can be set as follows: when the order response time reduction rate is less than 5%, the population size of the genetic algorithm is increased by 20%; when the pallet recycling rate declines for two consecutive cycles, the priority weight of the maintenance channel is increased.

[0072] Compared to existing technologies, current shared pallet rental solutions typically operate with fixed parameters after a single optimization, lacking continuous tracking and feedback adjustments to the actual results. For example, traditional systems do not reassess transportation cost changes based on subsequent order fulfillment data after route optimization, causing the optimization effect to decay over time. This solution, however, establishes a regular evaluation mechanism to identify optimization deviations caused by supply chain fluctuations and inventory allocation imbalances due to surges in holiday orders, thereby triggering targeted adjustments to algorithm parameters and creating a positive cycle of continuous optimization.

[0073] Through the aforementioned technical means, this invention solves the problem of insufficient adaptability of solutions caused by the rigidity of optimization parameters in existing technologies. By quantitatively evaluating actual operational effects and dynamically adjusting algorithm parameters, it ensures that the optimization strategy always matches the real-time business status. For example, when a new warehouse node is added to a region, the evaluation mechanism can quickly identify the inventory turnover rate data of that node and adjust the inventory allocation constraints of the genetic algorithm accordingly. This closed-loop feedback mechanism effectively maintains the continuity of optimization results and avoids solution failure due to environmental changes.

[0074] like Figure 2 As shown, this embodiment also proposes a shared pallet rental scheme optimization system, including: a data acquisition and preprocessing module 10, a digital twin modeling module 20, a multi-objective optimization module 30, and a function execution module 40.

[0075] like Figure 3 As shown, the function execution module 40 includes a rental pricing dynamic adjustment unit 401, an order intelligent matching unit 402, a recycling optimization unit 403, an abnormal order intelligent processing unit 404, and an optimization effect evaluation unit 405.

[0076] Among them, the data acquisition and preprocessing module 10 refers to the edge computing node deployed in the warehouse or outlet, which collects pallet entry and exit data, equipment status data and environmental data through sensors and 5G communication modules. Specifically, an industrial-grade IoT gateway can be used to realize data deduplication and outlier filtering, solving the problem of low efficiency of traditional manual collection.

[0077] The digital twin modeling module 20 refers to binding RFID tags to a virtual model and predicting pallet life through time series analysis. Specifically, Apache Kafka can be used to achieve real-time data stream synchronization, solving the problem of disconnect between physical assets and digital information.

[0078] The multi-objective optimization module 30 refers to a genetic algorithm using real-number encoding, whose adaptive crossover probability is dynamically adjusted through the number of generations. Specifically, the NSGA-II framework can be used to balance transportation costs and inventory turnover, solving the local optimum problem caused by fixed parameters.

[0079] The rental pricing dynamic adjustment unit 401 refers to the pricing factor trained based on the gradient boosting regression model. Specifically, the XGBoost algorithm can be used to process customer credit rating and rental period variables, thus solving the problem of the single pricing strategy in traditional pricing strategies.

[0080] The intelligent order matching unit 402 refers to the selection of the optimal network points by combining geographic information system and inventory availability. Specifically, the Dijkstra algorithm can be used to calculate the delivery route to solve the problem of response delay in manual matching.

[0081] Specifically, the data acquisition and preprocessing module 10 inputs the cleaned inventory data into the digital twin modeling module 20 to generate a virtual model that includes the degree of pallet wear and location coordinates.

[0082] The multi-objective optimization module 30 receives pallet status data output by the model and generates scheduling paths and pricing strategies through an improved genetic algorithm. The intelligent order matching unit 401 in the function execution module 40 prioritizes warehouses closest to customers and meeting availability requirements based on the optimization results, while simultaneously invoking the recycling optimization unit 403 to perform cluster analysis on the returned pallets and plan recycling routes. The intelligent abnormal order processing unit 404 monitors order status in real time using a random forest model, automatically triggering a financial reminder process when a settlement anomaly is detected. The optimization effect evaluation unit 405 periodically compares indicators such as inventory turnover rate and feeds feedback to the multi-objective optimization module for parameter iteration.

[0083] Compared to existing technologies, traditional systems rely on manual experience to adjust leasing strategies, failing to respond to real-time inventory changes. Existing technologies use fixed genetic algorithm parameters, leading to unstable optimization results, while this solution dynamically balances search efficiency through adaptive crossover probabilities. Existing recycling processes do not consider geographic clustering, while this solution uses the K-means algorithm to merge adjacent recycling points, reducing empty mileage. Existing pricing models do not integrate customer credit data, while this solution achieves differentiated pricing through machine learning models.

[0084] Through the aforementioned technical means, dynamic optimization of pallet scheduling routes was achieved, reducing the empty load rate of transport vehicles. Synchronizing physical pallet status using a digital twin model avoided inventory mismatches caused by information delays. A machine learning-driven anomaly detection mechanism shortened order anomaly response time. An intelligent matching strategy based on geographic location and inventory availability improved order processing efficiency. A closed-loop optimization evaluation mechanism ensured continuous optimization of system parameters, addressing the problem of insufficient iterative capabilities in traditional systems.

[0085] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope described in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing a shared pallet rental scheme, characterized in that, Includes the following steps: Acquire historical leasing data, real-time inventory data, and forecasted demand data for each node in the supply chain; Based on RFID technology, the entire lifecycle trajectory data of pallets is collected to establish a digital twin model of pallet circulation, which maps the entire process status of pallets from warehousing, leasing, transportation, recycling to repair / scrapping; The improved genetic algorithm is used to perform multi-objective optimization of pallet scheduling paths, inventory distribution, and rental pricing. In the improved genetic algorithm optimization step, the encoding method adopts real number encoding, and the pallet scheduling path, inventory allocation ratio and pricing coefficient are used as gene segments; The fitness function is constructed as follows: transportation cost weight coefficient × transportation cost value + inventory turnover rate weight coefficient × (1 - inventory turnover cycle) + customer satisfaction weight coefficient × customer satisfaction value. The weight coefficients are determined by the analytic hierarchy process. Adaptive crossover and mutation probabilities are introduced into genetic operations and dynamically adjusted according to the number of generations of population evolution to avoid the algorithm getting trapped in local optima; It also includes a dynamic adjustment step for lease pricing: a pricing factor system is built based on customer credit rating, lease volume, lease period and historical cooperation frequency. Customer credit rating is comprehensively assessed through account performance records, settlement timeliness and pallet loss rate. The pricing factors are trained using a gradient boosting regression model to generate personalized pricing schemes. When the customer's rental volume exceeds the second preset threshold or the rental period covers the industry's off-season, the discount coefficient is automatically adjusted. It also includes the optimization effect evaluation step: establishing an evaluation index system, which includes cost reduction rate, inventory turnover improvement rate, order response time reduction rate, customer complaint rate, and pallet recycling rate; Regularly compare and analyze the optimized actual operating data with the baseline data before optimization, and dynamically adjust the parameters of the genetic algorithm and the optimization target weights based on the evaluation results to achieve continuous iterative optimization of the leasing scheme.

2. The method for optimizing a shared pallet rental scheme according to claim 1, characterized in that, In the process of acquiring data from each node of the supply chain, edge computing nodes are deployed at warehouses / outlets to collect real-time data on pallet entry and exit, equipment operating status, and environmental temperature and humidity. The pre-processed data is uploaded to the cloud platform via a 5G communication module. The pre-processing process includes data deduplication, outlier removal, and format standardization. Outlier removal is based on the 3σ principle to identify loss rates and inventory quantity data that exceed the normal range.

3. The method for optimizing a shared pallet rental scheme according to claim 1, characterized in that, In the step of establishing a digital twin model of pallet circulation, the unique pallet identifier collected by RFID is bound to the virtual pallet in the digital twin model, and the physical status parameters of the pallet are synchronized in real time. The physical state parameters include wear level, load history, location coordinates and associated order information. The remaining service life of the pallet is predicted through time series analysis. When the remaining service life is lower than the first preset threshold, a maintenance reminder or scrap warning is automatically triggered.

4. The method for optimizing a shared pallet rental scheme according to claim 1, characterized in that, This includes an intelligent order matching process: based on the pallet type, quantity, delivery address, and usage period requirements in customer orders, combined with optimized inventory distribution data, it automatically matches the optimal outbound warehouse / outlet and carrier; During the matching process, priority is given to warehouses / outlets that are closest to the customer and have pre-set pallet availability, while also taking into account the carrier's historical on-time delivery rate and transportation cost quotes.

5. The method for optimizing a shared pallet rental scheme according to claim 1, characterized in that, The recycling optimization steps include: planning the optimal recycling route based on recycling application data initiated by members, the geographical distribution of recycling points, and pallet status data; and using clustering algorithms to merge adjacent recycling requests to reduce the empty mileage of recycling vehicles. The recycling pallets are graded and tested, and then allocated to the reuse, repair or disposal channels according to the test results. The test indicators include appearance integrity, structural strength and RFID signal stability.

6. The method for optimizing a shared pallet rental scheme according to claim 1, characterized in that, This includes intelligent processing steps for abnormal orders: real-time monitoring of rental order status through machine learning models to identify order delays, pallet damage, and settlement anomalies; The machine learning model is generated using the random forest algorithm, trained based on historical abnormal order data. Corresponding handling strategies are implemented for different types of abnormal triggers. When an order is delayed, the subsequent scheduling plan is automatically adjusted. When a pallet is damaged, the claims process is initiated and a replacement pallet is provided. When a settlement is abnormal, a reminder message is pushed to the finance module.

7. A shared pallet rental scheme optimization system, used to execute a shared pallet rental scheme optimization method according to any one of claims 1-6, characterized in that, include: The data acquisition and preprocessing module is used to acquire historical leasing data, real-time inventory data, and forecasted demand data from each node of the supply chain. The digital twin modeling module, based on the full lifecycle trajectory data of the pallet collected by RFID technology, establishes a digital twin model of pallet circulation, binds the unique pallet identifier collected by RFID to the virtual pallet in the digital twin model, and synchronizes the physical status parameters of the pallet in real time and maps the status of the entire process. The multi-objective optimization module uses an improved genetic algorithm to perform multi-objective optimization of pallet scheduling paths, inventory distribution, and rental pricing; It also includes functional execution modules, such as a rental pricing dynamic adjustment unit, an order intelligent matching unit, a recycling optimization unit, an abnormal order intelligent processing unit, and an optimization effect evaluation unit.

Images