Cloud-side collaborative logistics vehicle scheduling system and method fusing digital twinning

Abstract

The invention discloses a cloud-side collaborative logistics vehicle scheduling system and method fusing digital twinning, and relates to the technical field of vehicle scheduling, the system comprises a digital twinning module, a cloud platform module and an edge node module, and the modules carry out data interaction with edge computing equipment through a cellular network; the digital twin module comprises a scene dynamic modeling unit, an uncertainty simulation unit and a real-time mapping unit; the cloud platform module comprises a global optimization unit and a data processing unit; the edge node module comprises a real-time data acquisition unit, a data preprocessing unit, an instruction execution unit and a local decision unit; the method has the beneficial effects that digital twinning and cloud edge collaborative fusion is realized, and double guarantee of risk pre-judgment and real-time response is realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of vehicle scheduling, in particular to a cloud-edge collaborative logistics vehicle scheduling system and method fusing digital twinning. BACKGROUND

[0002] Logistics vehicle scheduling is the core link of logistics supply chain efficiency improvement. With the rapid development of e-commerce retail and intelligent manufacturing, the logistics scene presents three characteristics of order dynamicization, road condition complexification and vehicle model equipment diversification. The traditional static scheduling technology cannot meet the efficient, real-time and reliable operation requirements. In recent years, the logistics scheduling field has gradually introduced front-line technologies to solve industry needs, using cloud-edge collaboration technology to solve data transmission delay, using digital twinning technology to realize virtual mapping of physical scenes, using artificial intelligence algorithms to optimize decision-making strategies, and promoting the scheduling mode from passive adjustment to active prediction. However, the existing technology has low integration, and has not formed a complete solution covering risk prediction, real-time decision-making and closed-loop optimization.

[0003] The current logistics vehicle scheduling technology has the problem of poor scene adaptability in uncertain scenarios. The system usually uses a static model, which is difficult to cope with dynamic changes. And the existing scheduling model is based on the static assumption of fixed orders, stable road conditions and equipment failure. It does not establish a real-time mapping mechanism with the physical scene and cannot obtain multi-source dynamic data in real time. This leads to the deviation between model parameters and actual scene expanding over time, and the decision-making failure probability is high when facing unexpected situations. In addition, the existing technology has low processing efficiency for uncertain events, and is mostly a passive mode of adjusting after the occurrence of unexpected events, lacking the ability to predict and avoid risks. The reason is that the uncertainty simulation technology is not fused, and risk samples cannot be generated based on historical data and real-time trends, leading to scheduling adjustment lagging behind scene changes, large consumption of computing resources, slow response speed and high risk of order task delay. SUMMARY

[0004] The purpose of the present application is to solve the problem in the prior art that the real-time mapping mechanism with the physical scene is not established, the multi-source dynamic data cannot be obtained in real time, the deviation between the model parameters and the actual scene expands over time, and the decision-making failure probability is high when facing unexpected situations. A logistics vehicle scheduling system and method fusing digital twinning are proposed. The advantages of the present application are that digital twinning and cloud-edge collaboration are fused to realize the dual protection of risk prediction and real-time response.

[0005] To achieve the above object, the technical scheme adopted by the present application is: a cloud-edge collaborative logistics vehicle scheduling system fusing digital twinning, comprising a digital twinning module, a cloud platform module and an edge node module, data interaction between the modules being carried out through a cellular network and an edge computing device; the digital twinning module comprises a scene dynamic modeling unit, an uncertainty simulation unit and a real-time mapping unit; the cloud platform module comprises a global optimization unit and a data processing unit; the edge node module comprises a real-time data acquisition unit, a data preprocessing unit, an instruction execution unit and a local decision unit.

[0006] Preferably, the scene dynamic modeling unit constructs a digital mirror image of a physical scene based on map data, logistics vehicle static parameters, logistics vehicle dynamic data, energy supplement station configuration, station equipment information and dynamic road condition information, and realizes full mapping of scene elements; the uncertainty simulation unit generates dynamic risk samples by combining historical data using a Monte Carlo algorithm, and provides risk prediction basis for decision-making; the real-time mapping unit updates data at a second level using a dynamic parameter correction algorithm to adjust the digital twinning scene based on physical data collected by the edge node.

[0007] Preferably, the scheduling method comprises the following steps: S1, initializing a digital twinning scene, scheduling trigger conditions being: risk sample generation, data deviation exceeding a threshold value, execution not meeting the standard, and a sudden emergency event, any one of the scheduling conditions being triggered; S2, digital twinning synchronous mapping: importing initial data in the scene dynamic modeling unit, collecting real-time data by the real-time data acquisition unit in the edge node module, and updating the twinning scene by the real-time mapping unit; S3, risk simulation and global optimization: the uncertainty simulation unit outputs risk samples; the global optimization unit trains a model using reinforcement learning; the local decision unit outputs a global strategy, including path planning, energy supplement time period and order priority; S4, operation coordination synchronous execution: the edge node module preprocesses data, the data preprocessing unit filters GNSS positioning noise using a Kalman filtering algorithm, and parameters are distilled to protect privacy; when a sudden emergency event is triggered, a sudden emergency event processing sub-process is entered; when a sudden emergency event is not triggered, the local decision unit checks for no abnormalities, and issues an instruction to a vehicle-mounted terminal for execution, and the vehicle-mounted terminal executes the instruction; S5, feedback iteration optimization: the vehicle-mounted terminal returns an execution result, and the digital twinning module evaluates the effect by comparing a preset threshold value; S6, if the evaluation effect in step S5 meets the standard, the scheduling ends, and if the evaluation effect does not meet the standard, step S3 is returned to.

[0008] Preferably, the burst emergency event processing sub-process in step S4 includes the following steps: S41, sub-process entrance: the edge node module monitors burst data anomalies; S42, identify and classify the burst emergency event, call the local database, divide the burst emergency event into four levels, judge whether the local can be independently processed, the judgment conditions include the event level and whether the local processing time limit is met; S43, when the burst event level and the local processing time limit are both met, the local emergency strategy is generated; when there is a non-meeting in the burst event level or the local processing time limit, the cloud-edge collaborative supplementary processing is performed; the cloud-edge collaborative supplementary processing includes temporary disposal of the edge node module, and then the cloud end performs the corresponding; S44, local instruction execution and resource latching; S45, result feedback and algorithm iteration: upload the processing result to the digital twin module and evaluate the delay rate, if the delay rate is greater than 5%, optimize the emergency algorithm parameters; S46, sub-process exit: return to step S5.

[0009] Preferably, the four levels of the burst emergency event include vehicle failure as the first level, energy supply failure as the second level, temporary traffic control as the second level, emergency order insertion as the third level, and extreme weather in the whole region as the fourth level; the burst emergency event is less than or equal to the second level and the local processing time limit is 0-10 seconds, that is, the local emergency strategy is triggered, otherwise the cloud-edge collaborative supplementary processing is performed.

[0010] Preferably, the Monte Carlo algorithm is used for the digital twin module to generate risk samples: , wherein m is the historical data sample size; x k,t is the failure state of the kth data in the time window t; is an indicator function, which is 1 if it is true, and 0 otherwise; w k is the data weight, the recent data weight is higher, , , T is the current time.

[0011] Preferably, the Kalman filter algorithm is used for the edge node layer to correct the vehicle positioning data: prediction step: ; update step: ; scene adaptation: A (state transition matrix): ( seconds, uniform motion model); R (observation noise covariance): (positioning error, unit: m²).

[0012] Preferably, the global optimization unit is based on a reinforcement learning algorithm, and a reinforcement learning reward function is used to output an optimal scheduling strategy by the cloud platform layer: ; sub-reward item definition: (turnover rate reward, ). (latency reward, ); (cost reward, ).

[0013] Preferably, the dynamic parameter correction algorithm is used for the digital twin module to determine whether to trigger scheduling: ; trigger condition: ( , preset deviation threshold); weight distribution: vehicle endurance ( ), road congestion index ( ), energy supplement station status ( ), order quantity ( ).

[0014] Compared with the prior art, the application has the beneficial effects that: (1) Digital twin and cloud edge are cooperatively integrated to realize double protection of risk prediction and real-time response. Breakthrough the limitations of static modeling and passive response in the prior art, simulate uncertain risks in advance through digital twin, optimize strategies in advance through cloud, and avoid chaos caused by temporary adjustment after the event occurs; edge nodes focus on real-time task processing, solving the contradiction between global optimization delay and local decision blindness.

[0015] (2) Privacy protection, balancing sensitive data security and scheduling efficiency. In view of the core contradiction between data sharing and privacy leakage in the prior art, parameter distillation technology is adopted to realize data cooperation between cloud and edge nodes without transmitting order details, vehicle scheduling logs and other original sensitive data; compared with the traditional data transmission mode, it not only avoids the risk of privacy leakage, but also reduces the amount of data transmission and improves the cooperation efficiency.

[0016] (3) Closed-loop iterative optimization, adapting to dynamic scenarios to continuously improve scheduling performance. Unlike the defects of one-time decision based on static data in the prior art, through the feedback data of the vehicle terminal, the digital twin module continuously corrects the scene parameters and algorithm model, and the scheduling strategy is dynamically adjusted with the change of the operation scene. BRIEF DESCRIPTION OF DRAWINGS

[0017] Fig. 1 is the main flowchart of the application.

[0018] Fig. 2 is a sub-flowchart of the application. DETAILED DESCRIPTION

[0019] In the following, the application will be further described in conjunction with specific embodiments. It should be noted that, without conflict, the embodiments described below or the technical features between them can be combined to form new embodiments.

[0020] In the description of the present application, it should be noted that for orientation words, such as the terms "center", "transverse", "longitudinal", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. Indicate the orientation and positional relationship based on the orientation or positional relationship shown in the drawings, only for the convenience of describing the present application and simplifying the description, and cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and cannot be understood as limiting the specific protection scope of the present application.

[0021] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence.

[0022] Embodiments One of the preferred embodiments of the present application is shown as Figs. 1-2 A cloud-edge collaborative logistics vehicle scheduling system and method based on digital twinning, comprising a digital twinning module, a cloud platform module and an edge node module, each module interacts with edge computing devices through cellular network and edge computing devices; The digital twinning module comprises a scene dynamic modeling unit, an uncertainty simulation unit and a real-time mapping unit; The cloud platform module comprises a global optimization unit and a data processing unit; The edge node module comprises a real-time data acquisition unit, a data preprocessing unit, an instruction execution unit and a local decision unit.

[0023] Among them, taking the reverse short transportation scene as an example: the reverse short transportation scene has the core characteristics of short-distance high-frequency round trip, fixed loading and unloading point but dynamic demand, and vehicle turnover efficiency priority, which is easily affected by uncertainties such as queuing at loading and unloading points, temporary traffic control in the yard, and order insertion; Deployment of edge node module: 3 types of roadside edge nodes are deployed at the core path nodes of reverse short transportation, respectively located at: the entrance of the logistics yard, monitoring the vehicle driving out efficiency and road congestion in the park; Loading and unloading point, monitoring the idle state, queue length, etc.; Intersection on the way of transportation route. Deploy edge nodes at energy supplement stations to collect real-time idle, occupied, and fault states of the pile; Get the remaining power, charging time, and estimated completion time of the charging vehicle; Identify the number of queuing vehicles at the energy supplement station. Choose 15-20 pure electric heavy-duty trucks, each equipped with a Beidou dual-mode positioning system, a truck weight sensor, a power take-off state or a loading and unloading cargo door state sensor; Deployment of digital twinning module: based on a public cloud lightweight application server, accurately restore the layout of logistics yards, loading and unloading areas, energy supplement piles, driving or queuing routes, and vehicle models; Uncertainty simulation unit: using Python's Monte Carlo library, based on the short reverse transportation history data in the past half year, set the risk sample generation rule; Global optimization unit algorithm: reinforcement learning algorithm optimizes target weight adjustment, considers vehicle turnover rate, loading and unloading waiting time, charging cost and empty running rate, and adapts to the high frequency turnover demand of short reverse transportation; Edge node module algorithm: local decision unit embedded congestion fast bypass algorithm and multi-element energy supplement scheme planning algorithm in the yard, based on real-time traffic to quickly generate alternative routes.

[0024] The scene dynamic modeling unit constructs a digital mirror of the physical scene based on map data, logistics vehicle static parameters (vehicle configuration, loading capacity, battery capacity, etc.), logistics vehicle dynamic data (vehicle load, range, core system health status), energy supplement yard configuration (number of piles, fast and slow charging types, etc.), yard equipment information, dynamic traffic information, etc. Full mapping of scene elements is realized; The uncertainty simulation unit uses the Monte Carlo algorithm to generate dynamic risk samples by combining historical data (recent order mutation frequency, charging station failure probability, traffic congestion rules, etc.), providing risk prediction basis for decision-making; The real-time mapping unit updates the digital twin scene by using dynamic parameter correction algorithm to adjust the digital twin scene based on physical data collected by edge nodes, with second-level data updating, to ensure the real-time nature of the data.

[0025] The cloud platform module includes a global optimization unit and a data processing unit: Global optimization unit: based on reinforcement learning algorithm, input risk sample of digital twin module, regional order distribution, real-time load data of charging station, train scheduling model, output path planning, energy supplement time period allocation, order priority sorting global strategy.

[0026] Data processing unit: data cleaning, structuring and storage, scene parameter storage, providing data support for uncertainty simulation and algorithm iteration.

[0027] The edge node module includes a real-time data acquisition unit, a data preprocessing unit, an instruction execution unit, and a local decision unit: Real-time data acquisition unit: vehicle-mounted sensors, environmental sensors, dynamic traffic information, real-time state of energy supplement yard, vehicle operation data, and shipping task data.

[0028] Data preprocessing unit: uses Kalman filter algorithm to filter GNSS positioning noise, realizes vehicle-cloud data collaboration without transmitting original sensitive data, and avoids privacy leakage risk.

[0029] Instruction execution unit: receives cloud instructions to control vehicle navigation path, charging start and stop, and other set functions.

[0030] Local decision unit: receive cloud global strategy, start emergency adjustment algorithm for real-time burst situation, generate local correction scheduling instruction with second-level response, guarantee decision real-time performance.

[0031] Referring to Figs. 1-2 The scheduling method includes the following steps: S1, initialize digital twin scene, scheduling trigger conditions are: risk sample generation, data deviation threshold, execution substandard, emergency event, when any scheduling condition is met, it is triggered; S2, digital twin synchronization mapping: import initial data in scene dynamic modeling unit, including map, vehicle parameters, order; real-time data acquisition unit in edge node module collects real-time data, including road conditions, energy supplement state, vehicle endurance; real-time mapping unit updates twin scene (dynamic parameter correction); S3, risk simulation and global optimization: uncertainty simulation unit outputs risk sample; global optimization unit strengthens learning model; local decision unit outputs global strategy, including path planning, energy supplement period and order priority; S4, operation coordination synchronous execution: edge node module preprocesses data, data preprocessing unit uses Kalman filter algorithm to filter GNSS positioning noise, and parameter distillation protects privacy; when an emergency event is triggered, enter the emergency event processing sub-process; when an emergency event is not triggered, the local decision unit checks for no abnormalities, and issues an instruction to the vehicle terminal for execution; the vehicle terminal executes the instruction; S5, feedback iteration optimization: vehicle terminal returns execution results, including order completion, waiting time, and remaining endurance mileage; digital twin module evaluates the effect by comparing the preset threshold; S6, if the evaluation effect in step S5 meets the standard, the scheduling ends, otherwise, return to step S3.

[0032] Further, the emergency event processing sub-process in step S4 includes the following steps: S41, sub-process entry: edge node module monitors burst data anomaly; S42, identify and classify the emergency event (based on the level table in Table 1); call local database, including backup vehicle position within a specified range, adjacent energy supplement station state, and regional road network topology; divide the emergency event into four levels, judge whether the local can handle independently, and the judgment conditions include event level and whether the local processing time limit meets the standard; S43, when the emergency event level and the local processing time limit both meet the standard, the local emergency strategy is generated; when the emergency event level or the local processing time limit does not meet the standard, cloud-edge collaborative supplementary processing is performed; Table 1: The four levels of the emergency response are: vehicle malfunction (Level 1), energy replenishment operation malfunction (Level 2), temporary traffic control (Level 2), emergency order insertion (Level 3), and extreme weather across the region (Level 4). Local emergency response strategies include: Vehicle malfunction (Level 1): (1) Lock the malfunctioning vehicle task, (2) Dispatch a backup vehicle to take over, (3) Generate a rescue route; Refueling station failure (Level 2): ​​(1) Push failure information to surrounding vehicles; (2) Recommend the nearest alternative station with high availability; (3) Adjust vehicle routes. Temporary traffic control (Level 2): ​​(1) Generate detour routes, (2) Issue new navigation to affected vehicles, (3) Update traffic conditions in the twin module simultaneously; If the sudden emergency event is level 2 or lower and the local processing time is 0-10 seconds, the local emergency strategy will be triggered; otherwise, cloud-edge collaborative supplementary processing will be provided. Cloud-edge collaborative supplementary processing includes: Temporary handling of edge node modules: (1) Suspend tasks for affected vehicles, (2) Freeze related order allocations, (3) Transmit compressed data to the cloud; Cloud response (less than or equal to 30 seconds): (1) call the global capacity library, (2) reinforcement learning iteration strategy, (3) issue global collaborative instructions; S44, Local Instruction Execution and Resource Latching; S45. Result Feedback and Algorithm Iteration: Upload the processing results to the digital twin module and evaluate the delay rate. If the delay rate is greater than 5%, optimize the emergency algorithm parameters. S46. Subprocess exit: Return to step S5.

[0033] This method summarizes a four-step closed-loop approach to achieve efficient scheduling in uncertain scenarios: (1) Digital twin synchronous mapping: The digital twin module constructs an initial virtual scene, and the edge nodes upload physical data in real time to dynamically correct scene parameters, ensuring that the digital twin scene is consistent with the actual operation status.

[0034] Digital twin modeling and dynamic data updates for the scenario: The digital twin module imports road network data from a logistics park's warehousing area and driving network, combining this data with electric logistics vehicle parameters, charging station configurations, and task orders to construct an initial virtual reverse-shortage scenario. Edge nodes collect real-time data, including the number of logistics vehicles entering and exiting, road congestion index, the number of available workstations and queuing vehicles in the loading and unloading area, and monitoring for temporary traffic control or road congestion at intermediate intersections. Simultaneously, the vehicle-side uploads dynamic information about the logistics vehicles. The real-time mapping unit receives this data and synchronously updates the digital twin scenario.

[0035] (2) Risk simulation and global optimization: The digital twin module generates uncertainty risk samples, and the cloud global optimization unit combines samples and regional data to output the optimal scheduling strategy through reinforcement learning algorithm.

[0036] Uncertainty simulation and global optimization: The uncertainty simulation unit of the digital twin module generates risk samples based on historical data to estimate the possible changes in vehicle demand, transportation capacity demand, and energy supplement demand in the future time period. The global optimization unit of the cloud platform layer receives the samples, combines the full data, and trains the scheduling model through algorithm to output the global strategy.

[0037] (3) Cloud-edge collaborative execution: The edge node receives the cloud strategy, combines the real-time collected road conditions and energy supplement station status for local adjustment, and issues instructions to the vehicle terminal for execution.

[0038] After the edge node receives the cloud global strategy, the local decision unit combines real-time data for secondary verification, confirms that no adjustment is needed, and the edge node executes the scheduling instructions by the vehicle terminal.

[0039] (4) Feedback and iterative optimization: The vehicle terminal feeds back the execution results, and the digital twin module evaluates the scheduling effect. If it does not reach the preset threshold, the cloud will re-optimize to continuously improve the adaptability of the strategy.

[0040] The vehicle terminal real-time returns the execution results, including whether the freight order is completed on time, the remaining battery level, loading and unloading waiting time, and other data. The digital twin module receives the feedback data and evaluates the scheduling effect. If a sudden situation occurs, the edge node will trigger adjustment within seconds to adjust the vehicle instructions to ensure continuous scheduling without interruption.

[0041] In addition: The Monte Carlo algorithm is used by the digital twin module to generate risk samples: ; wherein: m is the historical data sample size (such as the fault record of the energy supplement station in the past 6 months); x k,t is the fault state of the kth data in the time window t (such as the offline of the energy supplement device); is an indicator function, which is 1 if it is true, and 0 otherwise; w k is the data weight, with higher weight for recent data, , , T is the current time.

[0042] Further: The Kalman filter algorithm (to filter vehicle positioning data noise) is used by the edge node layer to correct vehicle positioning data: Prediction step: ; Update step: ; Scenario adaptation: A (state transition matrix): ( Second, uniform motion model); R (observation noise covariance): (positioning error, unit: m²).

[0043] The global optimization unit is based on a reinforcement learning algorithm, and a reinforcement learning reward function (multi-objective optimization) is used to output the optimal scheduling strategy for the cloud platform layer: ; Sub-reward item definition: (turnover rate reward, ); (waiting time reward, ); (cost reward, ); Applicable scenarios: short transportation scenarios, high-frequency turnover requirements, and weight can be adjusted as needed.

[0044] In addition, the dynamic parameter correction algorithm and the bias calculation formula are used by the digital twin module to determine whether to trigger scheduling: δ = i = 1 n w i ∙ x i − x i 2; Trigger condition: δ > δ th (δ th = 0.15, preset bias threshold); Weight distribution: vehicle endurance (w veh = 0.3), road congestion index (w road = 0.25), energy station status (w charge = 0.25), and order quantity (w order = 0.2).

[0045] Finally, scheduling effect evaluation: key indicator formula: used for feedback iteration optimization link to determine whether the scheduling meets the standards: (1) Order punctuality rate: (2) Vehicle empty running rate: (3) Emergency event handling rate: .

[0046] The foregoing describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-described embodiments, and the above-described embodiments and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A cloud-edge collaborative logistics vehicle scheduling system integrating digital twins, characterized in that, include: The system includes a digital twin module, a cloud platform module, and an edge node module. These modules interact with each other via a cellular network and edge computing devices. The digital twin module includes a scene dynamic modeling unit, an uncertainty simulation unit, and a real-time mapping unit; The cloud platform module includes a global optimization unit and a data processing unit; The edge node module includes a real-time data acquisition unit, a data preprocessing unit, an instruction execution unit, and a local decision-making unit.

2. The cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 1, characterized in that, The scene dynamic modeling unit constructs a digital mirror of the physical scene based on map data, static parameters of logistics vehicles, dynamic data of logistics vehicles, configuration of refueling stations, station equipment information, and dynamic road condition information, thereby achieving full mapping of scene elements. The uncertainty simulation unit uses the Monte Carlo algorithm and combines historical data to generate dynamic risk samples, providing a basis for risk prediction for decision-making. The real-time mapping unit updates the physical data collected by the edge nodes in seconds and uses a dynamic parameter correction algorithm to adjust the digital twin scene.

3. The scheduling method of a cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 2, characterized in that, Includes the following steps: S1. Initialize the digital twin scenario. The scheduling trigger conditions are: risk sample generation, data deviation exceeding the threshold, execution failure, and sudden emergency events. The scenario will be triggered when any one of the scheduling conditions is met. S2, Digital Twin Synchronous Mapping: Import initial data from the scene dynamic modeling unit, collect real-time data from the real-time data acquisition unit in the edge node module, and update the twin scene from the real-time mapping unit. S3, Risk Simulation and Global Optimization: The uncertainty simulation unit outputs risk samples; the global optimization unit trains the model using reinforcement learning; the local decision-making unit outputs global strategies, including path planning, refueling time, and order priority. S4. Operation and coordination are synchronized: the edge node module preprocesses the data, the data preprocessing unit uses the Kalman filter algorithm to filter GNSS positioning noise, and parameter distillation protects privacy; when a sudden emergency event is triggered, the sudden emergency event handling sub-process is entered. If no emergency event is triggered, the local decision-making unit verifies that there are no abnormalities, issues instructions to the vehicle terminal for execution, and the vehicle terminal executes the instructions. S5. Feedback Iterative Optimization: The vehicle terminal transmits the execution results back, and the digital twin module evaluates the effect by comparing with preset thresholds; S6. If the evaluation results in step S5 meet the standards, the scheduling ends; otherwise, return to step S3.

4. The scheduling method of a cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 3, characterized in that, The emergency response sub-process in step S4 includes the following steps: S41, Sub-process entry point: Edge node module monitors sudden data anomalies; S42. Identify and classify sudden emergency events, call the local database, divide the sudden emergency events into four levels, and determine whether the local area can handle them independently. The judgment conditions include the event level and whether the local handling time meets the standard. S43. When both the emergency level and local processing time meet the standards, a local emergency strategy is generated; when either the emergency level or local processing time fails to meet the standards, cloud-edge collaborative supplementary processing is implemented; cloud-edge collaborative supplementary processing includes temporary handling by edge node modules, followed by corresponding actions in the cloud. S44, Local Instruction Execution and Resource Latching; S45. Result Feedback and Algorithm Iteration: Upload the processing results to the digital twin module and evaluate the delay rate. If the delay rate is greater than 5%, optimize the emergency algorithm parameters. S46. Subprocess exit: Return to step S5.

5. A cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 4, characterized in that, The four levels of the emergency events include vehicle malfunction (Level 1), energy replenishment failure (Level 2), temporary traffic control (Level 2), emergency order insertion (Level 3), and extreme weather across the entire region (Level 4). If the emergency event is Level 2 or lower and the local processing time is 0-10 seconds, the local emergency strategy will be triggered; otherwise, cloud-edge collaborative supplementary processing will be provided.

6. A cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 5, characterized in that, The Monte Carlo algorithm is used by the digital twin module to generate risk samples: Where m is the historical data sample size; x k,t The fault status of the k-th data item within the time window t; This is an indicator function; it returns 1 if the condition is met, and 0 otherwise. w k Data weights are assigned, with more recent data receiving higher weights. , , T This is the current time.

7. A cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 6, characterized in that, The Kalman filter algorithm is used to correct vehicle positioning data at the edge node layer. Prediction Step: ; Update steps: ; Scene adaptation: A (state transition matrix): ( (seconds, uniform motion model); R (observation noise covariance): (Positioning error, unit: m²).

8. A cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 7, characterized in that, The global optimization unit is based on a reinforcement learning algorithm, and the reinforcement learning reward function is used by the cloud platform layer to output the optimal scheduling strategy. Sub-reward item definition: (Turnover rate bonus) ); (Waiting time reward) ); (Cost incentive, ).

9. A cloud-edge collaborative logistics vehicle scheduling system integrating digital twins as described in claim 5, characterized in that, The dynamic parameter correction algorithm is used by the digital twin module to determine whether scheduling is triggered. ; Triggering conditions: ( (Preset deviation threshold); Weighting: Vehicle range ( ), Road congestion index ( ), Refueling Station Status ( ), number of orders ( ).

Images