A reverse path planning system for a circulating packaging box combined with forward logistics

Abstract

The application relates to the technical field of logistics distribution, and particularly discloses a circulating packaging box reverse path planning system combined with forward logistics. The system comprises a forward logistics task analysis module, a reverse logistics demand sensing module, a space-time state coupling modeling module, a collaborative path optimization engine and a dynamic instruction issuing and executing module. By constructing a vehicle space-time state transition model and utilizing an optimization algorithm to plan a collaborative path for the vehicle to execute the mixed distribution and recycling tasks, the application can integrate transport capacity, reduce empty mileage, reduce comprehensive transportation cost, dynamically respond to execution deviation and improve the overall operation efficiency of the system.

Description

Technical Field

[0001] This invention belongs to the field of logistics and distribution technology, specifically relating to a reverse path planning system for reusable packaging boxes that combines forward logistics. Background Technology

[0002] In the field of logistics and supply chain management, reusable packaging boxes, as a type of reusable logistics carrier, are of great significance for reducing operating costs, improving resource utilization efficiency, and achieving green logistics goals through efficient recycling and redistribution. Reverse logistics path planning for reusable packaging boxes is a crucial factor determining their recycling efficiency and economic benefits.

[0003] In existing technologies, reverse route planning for reusable packaging boxes is typically conducted independently of the forward logistics network. Reverse logistics vehicles plan the optimal recycling route based solely on the demand information of the recycling points, ignoring the ongoing forward logistics delivery tasks. This fragmented planning model often results in vehicles having to drive empty back to the recycling starting point after completing forward deliveries, or being unable to utilize the forward logistics capacity resources while performing recycling tasks.

[0004] Existing technologies suffer from path redundancy and high vehicle empty load rates, which not only increase transportation costs and carbon emissions but also reduce the overall operational efficiency of the logistics system. How to collaboratively optimize the reverse recycling path of reusable packaging boxes with the forward logistics delivery path to achieve the integration and efficient utilization of two-way logistics resources has become a pressing technical challenge in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a reverse path planning system for reusable packaging boxes that combines forward logistics, so as to solve the problems of high vehicle empty load rate, path redundancy, increased transportation costs and low overall system operating efficiency caused by the separation of forward and reverse logistics path planning in the prior art.

[0006] This invention provides a reverse path planning system for reusable packaging boxes that combines forward logistics, comprising: The forward logistics task parsing module is used to receive and parse the forward logistics delivery plan from the upstream order management system, extract the geographical coordinates of the delivery destination, the estimated delivery time window, the total volume and weight of the goods to be delivered, and the vehicle identification and vehicle specifications of the vehicle performing the delivery task. Based on the vehicle identification and vehicle specifications, it retrieves the maximum load, maximum volume, current fuel or battery status, and driving speed characteristic curve of the corresponding vehicle from the vehicle attribute database. The reverse logistics demand sensing module is used to collect recycling demand information of reusable packaging boxes, obtain the geographical coordinates of recycling points, the quantity and specifications of packaging boxes to be recycled, the earliest time expected to be loaded for recycling, and the latest deadline for recycling. The spatiotemporal state coupling modeling module is used to construct a mathematical model describing the evolution of the spatiotemporal state of vehicles during forward delivery and reverse recycling processes; The collaborative path optimization engine is used to search for the vehicle task execution sequence that minimizes the total system cost in the state-space network graph generated by the spatiotemporal state coupling modeling module. The dynamic command issuance and execution module is used to convert the optimal route plan generated by the collaborative route optimization engine into executable commands and monitor the execution process. It generates dynamic navigation and task lists for each vehicle participating in collaborative transportation and issues them through the vehicle network terminal or driver's mobile device.

[0007] Preferably, the reverse logistics demand perception module has a built-in packaging box status verification unit, which is used to query the standard volume and weight of the packaging box according to its specifications and model, and to verify the rationality of the reported quantity based on historical data, so as to output a structured reverse logistics demand list.

[0008] Preferably, the mathematical model takes the vehicle performing the forward logistics task as the dynamic subject and defines a multi-dimensional state vector that includes the vehicle's real-time geographical location, current load and cargo volume, completed forward delivery task sequence, executed reverse recycling task sequence, and cumulative driving mileage and operation time since its inception. The spatiotemporal state coupling modeling module receives the initial delivery plan and vehicle initial state from the forward logistics task parsing module and the reverse demand list from the reverse logistics demand perception module, and predicts the new state of the vehicle after executing any candidate task through the state transition function. The state transition function comprehensively considers the travel time between two points based on real-time road conditions, the standard time for loading and unloading operations at the task point, and the changes in vehicle load and volume caused by task execution, so as to output a dynamically updated spatial state network diagram containing all future states of all vehicles.

[0009] Preferably, the total system cost is a weighted sum of the total vehicle travel distance cost, the total operation time cost, the penalty cost for violating the time window, and the opportunity loss cost for not completing the retrieval task. The collaborative path optimization engine uses a metaheuristic algorithm based on large-scale neighborhood search to solve the problem. The execution process of the metaheuristic algorithm includes: taking the independent completion of all forward delivery tasks by each vehicle as the initial solution. During the iterative optimization phase, a vehicle and a continuous sequence of tasks are randomly selected from the current solution and removed to form a task pool containing forward delivery tasks and reverse recycling tasks to be reinserted. Attempt to re-insert tasks from the task pool into the paths of all vehicles in a new order and combination. The insertion position must satisfy vehicle load and volume constraints, time window constraints for each task, and upper limit constraints for total vehicle operation time. Calculate the total system cost of the candidate solutions generated for each successful insertion, and use simulated annealing as the acceptance criterion; After a preset number of iterations, the system outputs the vehicle routing scheme with the lowest total cost.

[0010] Preferably, the reverse demand list is arranged in chronological order and indicates the vehicle's next target point, task type, goods or packaging box identification to be processed, and planned time window. The dynamic instruction issuance and execution module receives the vehicle's GPS location information and task completion confirmation signal in real time. When the actual progress of the vehicle deviates from the plan, the collaborative path optimization engine is triggered to perform local replanning starting from the vehicle's current real-time status, re-optimize the remaining unfinished tasks, and generate adjusted instructions for re-issuance.

[0011] Preferably, the collaborative path optimization engine uses a metaheuristic algorithm based on large-scale neighborhood search, and the neighborhood destruction operator adopts an adaptive strategy. The adaptive strategy dynamically adjusts the number of tasks removed each time based on the improvement in the quality of the solution during the iteration process. Specifically, the metaheuristic algorithm maintains the range of values ​​for the number of tasks to be removed, and increases the upper limit of the number of tasks to be removed to expand the search range when a better solution is not found after a preset number of iterations. When better solutions are frequently found recently, reduce the number of removal tasks to focus on fine-grained search within the neighborhood of the current high-quality solution.

[0012] Preferably, the state transition function in the spatiotemporal state coupling modeling module integrates real-time traffic flow data into its travel time prediction sub-function; The travel time prediction sub-function is connected to the city traffic big data platform through the application programming interface to obtain the average traffic speed of each road segment on the planned route in the current and future prediction periods. It then uses a method based on a weighted calculation of historical average speed and real-time speed to predict the travel time. The weight of the real-time speed increases linearly as the time approaches the current moment.

[0013] Preferably, the rationality verification logic of the packaging box status verification unit in the reverse logistics demand perception module is based on a statistical model constructed from historical data. The statistical model records the historical average and standard deviation of the number of boxes recycled in a single transaction at each recycling point for each type of packaging box. When the newly reported quantity to be recycled exceeds the range of plus or minus three standard deviations of the historical average, the verification unit marks the historical data as abnormal and temporarily stores it, while sending a review request to the person in charge of the recycling point. Demand data will only be included in the reverse logistics demand list after receiving manual confirmation or correction.

[0014] Preferably, the dynamic instruction issuance and execution module adopts an incremental optimization strategy when triggering local replanning; The incremental optimization strategy fixes the task sequence that has been determined to be completed and cannot be changed, and only re-optimizes the incomplete tasks that have not yet started and those that are currently being executed but will be affected later, and sets the upper limit of the search time for replanning to a fixed value. If a feasible and superior solution to the original remaining plan is found within the time limit, the new solution is adopted; if no solution is found within the time limit, the original remaining plan is maintained and an early warning is issued to the dispatcher.

[0015] Preferably, the system operates on a hierarchical decision-making framework; the hierarchical decision-making framework includes a strategic configuration layer, a tactical planning layer, and a real-time operation layer; The strategic configuration layer is responsible for setting the system's basic parameters and long-term rules, including various cost weighting coefficients, vehicle fleet size and model configuration, and cooperation time window agreements signed with service outlets; The tactical planning layer consists of a collaborative path optimization engine, which generates detailed collaborative path plans on a daily or shift basis. The real-time operation layer is dominated by the dynamic instruction issuance and execution module, which is responsible for handling dynamic events and instruction execution at the minute to hour level. The three layers exchange parameters and commands through data interfaces.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention, by constructing a forward logistics task parsing module and a reverse logistics demand perception module, achieves synchronous collection and structured processing of forward delivery information and reverse recycling demand information at the data level, laying a data foundation for their collaborative optimization. The spatiotemporal state coupling modeling module creatively abstracts the process of vehicles performing bidirectional tasks into a unified state transition network, thereby transforming the complex path coordination problem into an optimization problem that can be searched in the state space. This method fundamentally breaks down the planning barriers between forward and reverse logistics in terms of modeling.

[0017] 2. The collaborative path optimization engine and its improved metaheuristic algorithm employed in this invention can efficiently search for hybrid task sequences with better global costs within a vast solution space. This solution allows vehicles to load recyclable packaging boxes along the way to or from delivery points, reducing the mileage required for dedicated empty recycling trips, lowering fuel consumption, vehicle wear and carbon emissions, and increasing the overall load factor of a single trip. This achieves deep integration and efficient utilization of transportation resources in both time and space dimensions.

[0018] 3. The dynamic instruction issuance and execution module and incremental replanning strategy of this invention endow the system with strong robustness in dealing with uncertainties in the logistics field. The system no longer generates static and rigid plans, but forms a dynamic closed loop of "planning-execution-monitoring-replanning". When the actual execution deviates from the plan, the system can respond quickly and re-optimize the remaining route based on the latest vehicle status, thereby ensuring the continuous optimality or near-optimality of the entire collaborative transportation solution in a dynamic environment, and improving the stability and efficiency of logistics system operation.

[0019] 4. This invention balances reliability and manageability through the design of the verification unit in the reverse logistics demand perception module and the hierarchical decision-making framework. The data verification mechanism ensures input quality and avoids erroneous decisions caused by junk data; the hierarchical framework separates long-term strategy, medium-term planning, and short-term execution, making the responsibilities of each level of the system clear. This facilitates parameter adjustment and strategy optimization, and also ensures the agility of real-time operations, making the solution easy to deploy and apply in complex real-world logistics environments. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework of the collaborative path optimization engine in this invention; Figure 3 This is a logical flow diagram of the spatiotemporal state coupling modeling module in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the forward logistics task analysis module and the reverse logistics demand perception module in this invention. Detailed Implementation

[0021] Example 1: Reference Figures 1 to 4 This invention proposes a reverse path planning system for reusable packaging boxes that integrates forward logistics. The system includes a forward logistics task analysis module, a reverse logistics demand perception module, a spatiotemporal state coupling modeling module, a collaborative path optimization engine, and a dynamic command issuance and execution module. These modules interact through standardized data interfaces, forming a complete closed loop from data acquisition, state modeling, path optimization to command execution and feedback. The specific implementation methods of each module will be described layer by layer, detailing its internal structure, data processing logic, parameter sources, state transition mechanisms, and collaborative relationships with other modules.

[0022] The forward logistics task parsing module receives and parses forward logistics delivery plans from the upstream order management system in real time. This module continuously monitors the delivery task flow at a throughput rate of no less than 10 messages per second via an enterprise service bus or message queue protocol. When a new delivery task is received, the module immediately initiates the parsing process. This process first verifies the integrity and validity of the task message, including checking the message signature, timestamp validity, and missing fields.

[0023] If the verification passes, the key parameters of the task are extracted, including the latitude and longitude coordinates of the delivery destination, the estimated delivery time window, the total volume and weight of the goods to be delivered, and the vehicle identifier and vehicle specification code for executing the forward logistics task. The vehicle identifier is a unique string used to index the static and dynamic attributes of the corresponding vehicle in the vehicle attribute database. The vehicle specification code is mapped to a preset vehicle configuration table, which defines the standard volume limit, maximum load capacity, number of loading and unloading doors, height, and other physical parameters for different vehicle models.

[0024] Furthermore, the forward logistics task parsing module retrieves the vehicle's current status data from the vehicle attribute database based on the vehicle identifier. This vehicle attribute database employs a hybrid architecture combining in-memory caching and persistent storage to ensure low-latency response under high-frequency reads. The retrieved data includes: the vehicle's maximum permissible load, maximum available volume, current remaining fuel or battery percentage, and a speed-energy consumption characteristic curve fitted based on historical driving data. This speed-energy consumption characteristic curve uses speed as the independent variable and energy consumption per unit distance as the dependent variable, and is used for accurate estimation of energy costs in subsequent route optimization.

[0025] All the above parameters are timestamped and updated synchronously after each vehicle status update. The module merges the original delivery task with the vehicle status data to generate a structured forward task record. Its data structure includes fields such as task ID, target coordinates, time window, cargo volume, cargo weight, vehicle ID, maximum vehicle load capacity, maximum vehicle volume, current battery level, and speed characteristic curve pointer. This forward task record is encapsulated in standard JSON format and pushed to the spatiotemporal state coupling modeling module via an internal message channel as the basic input for constructing the initial state network.

[0026] Meanwhile, the reverse logistics demand sensing module is responsible for collecting real-time recycling demand information for reusable packaging boxes. The input sources for this module mainly include two types: one is the IoT sensor terminals deployed at various recycling points, and the other is data manually reported by point operators through a dedicated mobile terminal application. The IoT sensor terminals have built-in pressure sensors, image recognition modules, and wireless communication units, which can automatically identify the quantity, model, and stacking status of packaging boxes stacked in a designated area. The terminals upload snapshots of the collected status to this module every 5 minutes. The mobile terminal application provides a graphical interface, allowing operators to input the specifications, model, quantity, earliest loading time (in Unix timestamps), and latest recycling deadline of the packaging boxes to be recycled. Both types of input data are transmitted encrypted via HTTPS and accompanied by digital signatures to ensure the authenticity of the source.

[0027] After receiving the raw data, the reverse logistics demand sensing module first performs data cleaning and format standardization. For IoT terminal data, the module calls an image recognition result verification algorithm to compare the consistency between sensor readings and visual recognition results; if the difference exceeds a preset threshold (e.g., 10%), it is marked as suspicious data and temporarily stored. For manually reported data, the module checks the logical rationality of the time window (e.g., the earliest time must not be later than the latest time). After preliminary cleaning, the data enters the packaging box status verification unit for in-depth verification. This packaging box status verification unit maintains a historical recycling database indexed by packaging box specifications and models, recording the statistical distribution of the number of each model recycled in a single transaction at each recycling point over the past 90 days.

[0028] For each (location, model) combination, the database stores its historical average recycling quantity. with standard deviation When the newly reported recycling quantity satisfy When the data is deemed statistically abnormal by the verification unit, the system automatically sends a verification request to the mobile device of the person in charge of the recycling point, requiring them to confirm or correct the data within 15 minutes. If no response is received within the time limit, the request is discarded; if confirmed to be correct, it is retained and marked as "human verification passed"; if corrected, the corrected value is used. This mechanism filters out noisy data caused by sensor drift, accidental touches, or human input errors. The verified request data is constructed into a structured reverse task record, including fields such as the recycling point coordinates, packaging box model, quantity, earliest loading time, latest deadline, and verification status, and is pushed to the spatiotemporal state coupling modeling module.

[0029] The spatiotemporal state coupling modeling module serves as a bridge connecting data acquisition and path optimization; its logical flow is shown in the attached diagram. Figure 3As shown, this spatiotemporal state coupling modeling module constructs a dynamically evolving state-space network graph. Each node represents the complete state of a vehicle at a certain moment, and each directed edge represents the state transition after executing a task. During module initialization, it receives N forward task records from the forward logistics task parsing module and M reverse task records from the reverse logistics demand perception module. Simultaneously, the module obtains the initial states of all K vehicles participating in the operation that day from the vehicle dispatch center, including departure warehouse coordinates, initial load (usually 0), initial volume occupancy (usually 0), initial battery level (usually 100%), and departure time (usually 06:00 of the day).

[0030] Based on the above input, the module begins to construct the initial state network. For each vehicle... Its initial state Defined as a seven-dimensional vector: , Geographic coordinates For the current load, Current volume occupancy This is a list of completed positive task IDs. This is a list of reverse engineering task IDs that have been executed. This is the current system time (in seconds). Initially, and All are empty lists. and =0, This refers to the departure time.

[0031] Subsequently, the module defines the state transition function. Used to calculate from state Execute the task The new state after The execution logic of the state transition function defined in this module is as follows: Based on... coordinates in and target coordinates The driving time prediction sub-function is called to calculate the estimated driving time between the two points. This travel time prediction sub-function connects to the city's traffic big data platform via API to obtain the predicted travel speed of each road segment along the route at a 5-minute granularity for the next hour. (The remaining text appears to be incomplete and requires further context.) The weighting coefficient for the time period is: In other words, the closer to the current moment, the higher the weight of real-time data. The final travel time is calculated from the weighted average speed.

[0032] Determine the loading and unloading operation time based on the task type. For forward delivery, the delivery time is determined by the volume of the goods, as shown in the formula: , This is the loading and unloading coefficient per unit volume (default value is 120 seconds / cubic meter). The volume of goods for forward missions; for reverse missions, the time taken depends on the number of boxes, as shown in the formula. , This is the single-box processing coefficient (default value is 30 seconds / box). This represents the number of packaging boxes for the reverse task. Update load and volume: increase load and volume for forward tasks, decrease load but increase volume for reverse tasks (due to space occupied by recycling empty boxes). Finally, add the new task ID to the corresponding completed task list and update the cumulative mileage and operation time.

[0033] ; Number of segments for the path For the first Segment length, For the number of historical and real-time data sources, For the first Data sources at time offset The weight of the position, For the first The section in Predicted speeds from multiple data sources. The formula above illustrates the travel time prediction mechanism based on multi-source data fusion.

[0034] By repeatedly applying state transition functions, the module generates a large state network. Each node in the network carries a complete state vector, and each edge is labeled with the executed task ID, cost increment, and time consumption. This state network is not static but dynamically expands as new tasks are added or the state is updated. The module employs an incremental graph construction strategy, expanding new branches only when necessary to avoid memory explosion. The completed state network is serialized into an adjacency list format and passed to the cooperative path optimization engine.

[0035] The principle framework of the collaborative path optimization engine is attached. Figure 2 As shown. The goal of this cooperative path optimization engine is to search for a set of vehicle paths in the state network such that the total system cost is reduced. Minimum. The total system cost is defined as: ; Total distance traveled by all vehicles (in kilometers). Total operating time (hours) for all vehicles. The penalty items (in yuan) are for violating the time window. This represents the number of packaging boxes (in units) that have not yet completed their recycling tasks. , , , These are the corresponding cost weighting coefficients, which are pre-set by the strategic allocation layer. For example, It can be set to 2.5 yuan / km. It costs 50 yuan per hour. It costs 200 yuan per time. It costs 10 yuan per piece.

[0036] The engine employs an improved large-scale neighborhood search algorithm to solve the combinatorial optimization problem. In the initialization phase, each vehicle is assigned its original forward task sequence, ignoring all reverse tasks, thus forming an initial feasible solution. This is followed by an iterative optimization loop. In each iteration, the algorithm first determines the damage intensity based on an adaptive strategy. . The initial value is 3, indicating that 3 consecutive tasks are removed each time. The algorithm maintains a sliding window, recording the number of times a better solution is found in the last 20 iterations. .like If the search is trapped in a local optimum, then it is considered that the search has been trapped in a local optimum. The upper limit has been increased from 5 to 8; if Then the upper limit of p decreases from 5 to 3. The actual number of removals is... Randomly select from the range This represents the maximum destructive strength.

[0037] After selecting vehicles and task segments, the algorithm forms a task pool by combining these tasks with all unassigned reverse tasks. Then, using a greedy insertion heuristic, it attempts to insert each task in the task pool into all legal positions on all vehicle paths. Insertion validity requires simultaneously satisfying the following conditions: vehicle load not exceeding maximum load, volume occupancy not exceeding maximum volume, task arrival time falling within its time window, and total vehicle operation time not exceeding 12 hours (the legal limit). Each successful insertion generates a candidate solution, and the engine calculates its total cost. If the total cost is better than the current optimal solution, it is accepted; otherwise, simulated annealing probability is used. accept, This represents the total cost corresponding to the currently retained optimal solution. The current temperature decays exponentially with the number of iterations.

[0038] After 5000 iterations or 500 consecutive iterations without improvement, the algorithm terminates and outputs the optimal route. This optimal route specifies a mixed task sequence for each vehicle, including task type, target coordinates, cargo / packaging information, and planned arrival and departure times.

[0039] Upon receiving the optimal route plan, the dynamic command issuance and execution module immediately generates a dynamic navigation and task list for each vehicle. The list is pushed via the in-vehicle terminal or the driver's mobile app, with content arranged chronologically. Each task clearly indicates: the coordinates of the next destination point, the task type (delivery / recovery), a list of cargo batch numbers or packaging box numbers, and the planned time window. Simultaneously, the module initiates a real-time monitoring process, receiving the vehicle's GPS position, speed, and task completion confirmation signals every 10 seconds.

[0040] When a deviation from the actual progress is detected, the module triggers a local replanning. Replanning employs an incremental strategy: completed tasks are frozen and cannot be changed; currently executing tasks and their successors are included in the re-optimization scope. Replanning invokes a lightweight instance of the collaborative path optimization engine, optimizing only the affected sub-paths, with a strict computation time limit of 30 seconds. If a lower-cost alternative is found within this timeframe, a new instruction is immediately generated and issued; otherwise, the original plan is maintained, and an alert message is sent to the scheduling center console, prompting manual intervention.

[0041] The entire system operates within a hierarchical decision-making framework. The strategic configuration layer sets long-term parameters, such as cost weights, fleet size, and network cooperation time windows; the tactical planning layer (i.e., the collaborative route optimization engine) generates detailed routes on a shift-by-shift basis; and the real-time operation layer (i.e., the dynamic command issuance and execution module) handles minute-level dynamic events. These three layers are connected via RESTful... API parameter synchronization ensures strategy consistency. This hybrid architecture enables the system to adapt to both long-term business changes and short-term operational disruptions, achieving deep integration and efficient collaboration between reverse logistics and forward delivery of reusable packaging boxes.

[0042] Example 2: Building upon Example 1, this example further refines the neighborhood structure in the collaborative path optimization engine to improve search efficiency in high-density task scenarios. Specifically, when the number of reverse recycling tasks to be processed in the system is more than twice the number of forward delivery tasks, the traditional random removal strategy results in a large number of reverse tasks being repeatedly moved in and out, leading to a decrease in convergence speed. Therefore, this example introduces a task-type-aware neighborhood destruction operator.

[0043] In each iteration, the neighborhood destruction operator first calculates the ratio of forward to reverse tasks in the current solution. If the reverse task ratio is greater than 60%, it prioritizes removing continuous segments containing reverse tasks. Specifically, the algorithm traverses all vehicle paths, identifies all continuous subsequences containing at least one reverse task, and calculates their length and reverse task density (number of reverse tasks / total number of tasks). Then, using the reverse task density as a weight, it performs weighted random sampling, selecting a high-density reverse task sequence for removal. This ensures that in scenarios dominated by reverse tasks, the algorithm can focus more on optimizing the integration efficiency of the recovery paths.

[0044] This embodiment expands the state vector in the spatiotemporal state coupling modeling module by adding a "remaining effective working time of the vehicle" component. This component is derived by subtracting the accumulated working time from the vehicle's legally maximum working time (e.g., 12 hours) and is updated synchronously during state transitions. During path optimization, this "remaining effective working time of the vehicle" component is used as a hard constraint; any insertion operation that results in a negative remaining effective working time is directly rejected, thus avoiding the generation of infeasible solutions that violate labor regulations.

[0045] In the dynamic command issuance phase, this embodiment adds a multi-level early warning mechanism. When a local replanning fails to generate a new plan due to timeout, the system not only issues an early warning to the dispatcher but also activates different levels of contingency plans based on the degree of delay. For example, if the delay is between 15 and 30 minutes, the system automatically extends the time window for subsequent tasks; if the delay is greater than 30 minutes, it attempts to reassign some non-emergency recovery tasks to other nearby vehicles, provided that these vehicles have sufficient time and space leeway on their paths. This multi-level early warning mechanism further enhances the system's robustness under extreme disturbances.

[0046] Through the above enhancements, this embodiment improves the system's performance in complex scenarios with intensive reverse engineering tasks and strict time constraints while maintaining the original architecture, thus verifying the scalability and adaptability of the technical solution of this invention.

[0047] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0048] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A reverse path planning system for reusable packaging boxes that combines forward logistics, characterized in that, include: The forward logistics task parsing module is used to receive and parse the forward logistics delivery plan from the upstream order management system, extract the geographical coordinates of the delivery destination, the estimated delivery time window, the total volume and weight of the goods to be delivered, and the vehicle identification and vehicle specifications of the vehicle performing the delivery task. Based on the vehicle identification and vehicle specifications, it retrieves the maximum load, maximum volume, current fuel or battery status, and driving speed characteristic curve of the corresponding vehicle from the vehicle attribute database. The reverse logistics demand sensing module is used to collect recycling demand information of reusable packaging boxes, obtain the geographical coordinates of recycling points, the quantity and specifications of packaging boxes to be recycled, the earliest time expected to be loaded for recycling, and the latest deadline for recycling. The spatiotemporal state coupling modeling module is used to construct a mathematical model describing the evolution of the spatiotemporal state of a vehicle during forward delivery and reverse recycling processes. The collaborative path optimization engine is used to search for the vehicle task execution sequence that minimizes the total system cost in the state-space network graph generated by the spatiotemporal state coupling modeling module. The dynamic command issuance and execution module is used to convert the optimal route plan generated by the collaborative route optimization engine into executable commands and monitor the execution process. It generates dynamic navigation and task lists for each vehicle participating in collaborative transportation and issues them through the vehicle network terminal or driver's mobile device.

2. The reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 1, characterized in that, The reverse logistics demand perception module has a built-in packaging box status verification unit, which is used to query the standard volume and weight of the packaging box according to its specifications and model, and to verify the rationality of the reported quantity based on historical data, so as to output a structured reverse logistics demand list.

3. The reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 2, characterized in that, The mathematical model takes the vehicle performing the forward logistics task as the dynamic subject and defines a multi-dimensional state vector that includes the vehicle's real-time geographical location, current load and cargo volume, completed forward delivery task sequence, executed reverse recovery task sequence, and cumulative driving mileage and operation time since its inception. The spatiotemporal state coupling modeling module receives the initial delivery plan and vehicle initial state from the forward logistics task parsing module and the reverse demand list from the reverse logistics demand perception module, and predicts the new state of the vehicle after executing any candidate task through the state transition function. The state transition function comprehensively considers the travel time between two points based on real-time road conditions, the standard time for loading and unloading operations at the task point, and the changes in vehicle load and volume caused by task execution, so as to output a dynamically updated spatial state network diagram containing all future states of all vehicles.

4. The reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 3, characterized in that, The total system cost is a weighted sum of the total vehicle travel distance cost, the total operation time cost, the penalty cost for violating the time window, and the opportunity loss cost for not completing the retrieval task. The collaborative path optimization engine uses a metaheuristic algorithm based on large-scale neighborhood search to solve the problem. The execution process of the metaheuristic algorithm includes: taking the independent completion of all forward delivery tasks by each vehicle as the initial solution. During the iterative optimization phase, a vehicle and a continuous sequence of tasks are randomly selected from the current solution and removed to form a task pool containing forward delivery tasks and reverse recycling tasks to be reinserted. Attempt to re-insert tasks from the task pool into the paths of all vehicles in a new order and combination. The insertion position must satisfy vehicle load and volume constraints, time window constraints for each task, and upper limit constraints for total vehicle operation time. Calculate the total system cost of the candidate solutions generated for each successful insertion, and use simulated annealing as the acceptance criterion; After a preset number of iterations, the system outputs the vehicle routing scheme with the lowest total cost.

5. A reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 4, characterized in that, The reverse demand list is arranged in chronological order and indicates the vehicle's next target point, task type, pending goods or packaging box identification, and planned time window. The dynamic instruction issuance and execution module receives the vehicle's GPS location information and task completion confirmation signal in real time. When the vehicle's actual progress deviates from the plan, the collaborative path optimization engine is triggered to perform local replanning starting from the vehicle's current real-time status, re-optimize the remaining unfinished tasks, and generate adjusted instructions for re-issuance.

6. A reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 5, characterized in that, The collaborative path optimization engine uses a metaheuristic algorithm based on large-scale neighborhood search, and the neighborhood destruction operator adopts an adaptive strategy. The adaptive strategy dynamically adjusts the number of tasks removed each time based on the improvement in the quality of the solution during the iteration process; Specifically, the metaheuristic algorithm maintains a range of values ​​for the number of tasks to be removed. If a better solution is not found after a preset number of iterations, the upper limit of the number of tasks to be removed is increased to expand the search range. When better solutions are frequently found recently, reduce the number of removal tasks to focus on fine-grained search within the neighborhood of the current high-quality solution.

7. A reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 6, characterized in that, The state transition function in the spatiotemporal state coupling modeling module integrates real-time traffic flow data into its travel time prediction sub-function. The travel time prediction sub-function is connected to the city traffic big data platform through the application programming interface to obtain the average traffic speed of each road segment on the planned route in the current and future prediction periods. It then uses a method based on a weighted calculation of historical average speed and real-time speed to predict the travel time. The weight of the real-time speed increases linearly as the time approaches the current moment.

8. A reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 7, characterized in that, The rationality verification logic of the packaging box status verification unit in the reverse logistics demand perception module is based on a statistical model constructed from historical data. The statistical model records the historical average and standard deviation of the number of boxes recycled in a single transaction at each recycling point for each type of packaging box. When the newly reported quantity to be recycled exceeds the range of plus or minus three standard deviations of the historical average, the verification unit marks the historical data as abnormal and temporarily stores it, while sending a review request to the person in charge of the recycling point. Demand data will only be included in the reverse logistics demand list after receiving manual confirmation or correction.

9. A reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 8, characterized in that, When the dynamic instruction issuance and execution module triggers local replanning, it adopts an incremental optimization strategy. The incremental optimization strategy fixes the task sequence that has been determined to be completed and cannot be changed, and only re-optimizes the incomplete tasks that have not yet started and those that are currently being executed but will be affected later, and sets the upper limit of the search time for replanning to a fixed value. If a feasible and superior solution to the original remaining plan is found within the time limit, the new solution is adopted; if no solution is found within the time limit, the original remaining plan is maintained and an early warning is issued to the dispatcher.

10. A reverse path planning system for reusable packaging boxes combined with forward logistics according to claim 9, characterized in that, The system operates on a hierarchical decision-making framework, which includes a strategic configuration layer, a tactical planning layer, and a real-time operation layer. The strategic configuration layer is responsible for setting the system's basic parameters and long-term rules, including various cost weighting coefficients, vehicle fleet size and model configuration, and cooperation time window agreements signed with service outlets; The tactical planning layer consists of a collaborative path optimization engine, which generates detailed collaborative path plans on a daily or shift basis. The real-time operation layer is dominated by the dynamic instruction issuance and execution module, which is responsible for handling dynamic events and instruction execution at the minute to hour level. The three layers exchange parameters and commands through data interfaces.

Images