A multi-scene dining delivery intelligent scheduling management system

Abstract

The application discloses a multi-scene dining distribution intelligent scheduling management system and relates to the technical field of dining distribution intelligent scheduling.The system comprises a multi-source road condition fusion sensing module, a fusion effectiveness intelligent regulation and control module, a high-concurrency computing power evaluation module and a computing power adaptability intelligent regulation and control module.The multi-source road condition fusion sensing module acquires data and evaluates fusion effectiveness.The fusion effectiveness intelligent regulation and control module decides whether to regulate and control according to the fusion effectiveness, and the updated road condition information is transmitted to the high-concurrency computing power evaluation module.The high-concurrency computing power evaluation module evaluates the computing power adaptability, and the computing power adaptability intelligent regulation and control module decides whether to regulate and control according to the computing power adaptability.Finally, the distribution scheduling instruction is transmitted, and the reliability of the dining distribution intelligent scheduling management is improved.The problem of low reliability of the dining distribution intelligent scheduling management under the multi-scene and high-concurrency pressure in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology for food delivery, and in particular to an intelligent scheduling and management system for multi-scenario food delivery. Background Technology

[0002] In multi-scenario food delivery, firstly, natural language processing is used to parse order notes, and then GIS (Geographic Information System) maps and BIM (Building Information Modeling) models are combined to achieve accurate address location. At the same time, LSTM (Long Short-Term Memory) neural network prediction models are used to predict the heat distribution of orders in different scenarios such as campuses, office buildings, and communities. Merchants are linked to obtain real-time food preparation progress, and orders are digitally marked according to characteristics such as temperature zone demand and timeliness level.

[0003] Next, the dynamic order allocation algorithm based on reinforcement learning completes the optimal matching of orders and transportation capacity. The three-dimensional packing algorithm plans the food loading scheme and automatically allocates corresponding temperature zone equipment or large-capacity transportation capacity for special orders such as low-temperature food and group meals. With the help of the "cloud-edge-device" distributed architecture, the real-time scheduling of global transportation capacity is realized, and riders or unmanned equipment are pre-scheduled to the waiting area before the noon peak.

[0004] Subsequently, the dynamic route planning module, integrating ant colony algorithms and real-time traffic data, refreshes the globally optimal route once within a preset time. It generates circular relay or grid flow routes for scenarios such as complex campus building distribution and urban road congestion. At the same time, it simulates the delivery process in the cloud using digital twin technology, predicts potential capacity gaps, and makes adjustments in advance. During the fulfillment phase, non-urgent orders are prioritized for allocation to lockers, and the number of lockers opened is dynamically adjusted using a locker utilization heatmap. Meanwhile, IoT sensors monitor the temperature and humidity of the food and the status of the equipment in real time. If any abnormality occurs, a reinforcement learning-driven reassignment mechanism is immediately triggered. Finally, the data after delivery is completed is fed back and Monte Carlo tree search is used to optimize subsequent matching strategies.

[0005] The existing dispatch system relies on a combination of third-party map APIs (Application Programming Interfaces) and self-built monitoring systems for road condition data. Due to the lack of effective data time series calibration algorithms and accuracy adaptation models, it is impossible to uniformly normalize the timestamps and perception dimensions of the two types of data. This easily leads to the phenomenon of "main roads showing smooth traffic while internal branch roads are actually congested," causing planned routes to get stuck at cross-regional connection nodes. Riders are forced to temporarily deviate from the planned routes and find feasible paths again, which significantly increases the extra time and uncertainty of last-mile delivery.

[0006] Meanwhile, during peak order periods (such as lunch and dinner rush hours), there are usually hundreds of riders and dozens of unmanned delivery vehicles in a single area. Each route recalculation requires complex iterative calculations using massive amounts of data from multiple dimensions, including road conditions, order locations, real-time delivery capacity, and meal timeliness requirements. Because the allocation of computing resources cannot dynamically match the explosive growth in business volume, the computing capacity per unit time quickly reaches its bottleneck. Meanwhile, riders are in a continuous state of movement during the delivery process, and during this delay, they may have already reached the entrance of a congested section of road. Even if they receive new route instructions later, they cannot turn around or detour in time and can only wait in the congested section, resulting in a significant decrease in delivery timeliness. This presents a problem of low reliability in intelligent scheduling and management of meal delivery under multiple scenarios and high concurrency pressure. Summary of the Invention

[0007] To address the low reliability of existing intelligent scheduling and management systems for food delivery under multiple scenarios and high concurrency pressure, this invention provides a multi-scenario intelligent scheduling and management system for food delivery. The technical solution is as follows:

[0008] On one hand, a multi-scenario intelligent dispatch and management system for meal delivery is provided. This system includes a multi-source traffic condition fusion perception module, a fusion effectiveness intelligent control module, a high-concurrency computing power evaluation module, and a computing power adaptability intelligent control module. The multi-source traffic condition fusion perception module is used to import multi-source heterogeneous traffic condition data from third-party map APIs and self-built monitoring systems. Based on this multi-source heterogeneous traffic condition data, it obtains a multi-source traffic condition data fusion effectiveness evaluation, which reflects the quantitative degree of accuracy in cross-scenario traffic condition information synchronization and the effectiveness of route planning data. The fusion effectiveness intelligent control module is used to determine whether to perform fusion effectiveness intelligent control based on the multi-source traffic condition data fusion effectiveness evaluation. If so, it generates updated fusion traffic condition information after control and transmits it to the high-concurrency computing power evaluation module; otherwise, it directly generates updated information. The merged traffic information is then transmitted to the high-concurrency computing power assessment module. The intelligent control of fusion effectiveness includes time granularity control and cross-validation node number control. The high-concurrency computing power assessment module is used to acquire high-concurrency computing power data in high-concurrency scenarios. Based on the high-concurrency computing power data, a high-concurrency scenario computing power adaptability assessment is obtained to reflect the quantitative degree of adaptability between delivery timeliness and intelligent scheduling of meal delivery in high-concurrency scenarios. The computing power adaptability intelligent control module is used to determine whether to execute intelligent computing power adaptability control based on the high-concurrency scenario computing power adaptability assessment. If yes, an intelligent scheduling instruction for meal delivery is sent after control. If no, an intelligent scheduling instruction for meal delivery is sent directly. The execution of intelligent computing power adaptability control includes low-priority instruction cache queue length threshold control and collaborative task splitting granularity control.

[0009] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0010] 1. By integrating heterogeneous data from third-party map APIs and self-built monitoring systems, and combining the fusion effectiveness assessment to quantify the accuracy of cross-scenario traffic condition synchronization and the effectiveness of route planning data, a high-quality data foundation is provided for subsequent scheduling decisions. The fusion effectiveness intelligent control module further optimizes the data fusion strategy dynamically through time granularity control and cross-validation node number control, effectively compensating for frequency differences and spatial deviations in multi-source data. This ensures that the generated fused traffic information can accurately match the traffic characteristics of different scenarios such as campuses, business districts, and communities, avoiding unreasonable route planning due to data errors, and providing reliable environmental awareness support for delivery scheduling.

[0011] 2. By accurately acquiring computing power data and assessing computing power adaptability for scenarios such as peak order periods, the system clearly understands the matching status between delivery timeliness and scheduling requirements. Based on the assessment results, the intelligent computing power adaptability control module adjusts the low-priority instruction cache queue length threshold to prioritize the release of computing power to core scheduling tasks during high concurrency. Through granular control of collaborative task splitting, it balances parallel processing efficiency and communication costs, avoiding scheduling delays caused by computing power waste or resource contention. This dynamic control mechanism ensures that computing power resources can be tilted towards critical tasks under high-pressure scenarios such as order surges, guaranteeing low latency and high availability of scheduling instruction processing.

[0012] 3. By establishing a closed-loop collaborative mechanism of data fusion, effectiveness regulation, computing power assessment, and computing power regulation, the entire process from traffic data processing to dispatch command issuance is made intelligent. On the one hand, accurate fusion of traffic information provides a scientific basis for initial route planning; on the other hand, dynamic adaptation of computing power for high-concurrency scenarios ensures that dispatch commands can respond promptly to real-time changes in orders, transportation capacity, and traffic conditions. Whether it's concentrated ordering during school breaks, delivery during lunch and evening peak hours in shopping districts, or scattered order dispatching in communities, the optimal dispatch plan can be quickly generated through the collaborative operation between modules, effectively reducing delivery delay rates and improving overall delivery efficiency and user dining experience. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of the structure of a multi-scenario intelligent dispatch and management system for meal delivery provided in an embodiment of the present invention;

[0015] Figure 2 This is a flowchart illustrating the cross-validation node quantity control of a multi-scenario intelligent dispatch and management system for meal delivery provided in an embodiment of the present invention.

[0016] Figure 3 This is a flowchart illustrating the collaborative task breakdown granularity control of a multi-scenario intelligent scheduling and management system for meal delivery provided in an embodiment of the present invention. Detailed Implementation

[0017] This application provides a multi-scenario intelligent dispatch and management system for food delivery, which solves the problem of low reliability in existing intelligent dispatch and management systems for food delivery under multiple scenarios and high concurrency pressure. The system acquires data and evaluates the effectiveness of the fusion through a multi-source road condition fusion perception module; the intelligent control module for fusion effectiveness decides whether to control based on this, and generates updated road condition information which is transmitted to the high-concurrency computing power evaluation module; the latter evaluates the computing power adaptability, and the intelligent control module for computing power adaptability decides whether to control, and finally sends delivery dispatch instructions, thereby improving the reliability of intelligent dispatch and management of food delivery.

[0018] The technical solution in this application aims to address the low reliability of intelligent scheduling and management of meal delivery under multiple scenarios and high concurrency pressure. The overall approach is as follows:

[0019] By importing multi-source heterogeneous traffic condition data from third-party map APIs and a self-built monitoring system, an effectiveness assessment of multi-source traffic condition data fusion is obtained. Based on the effectiveness assessment, it is determined whether to implement intelligent control based on fusion effectiveness. If so, updated fused traffic condition information is generated after control and transmitted to the high-concurrency computing power assessment module. If not, updated fused traffic condition information is directly generated and transmitted to the high-concurrency computing power assessment module to obtain high-concurrency computing power data under high-concurrency scenarios. Based on the high-concurrency computing power data, a computing power adaptability assessment for high-concurrency scenarios is obtained. Based on the computing power adaptability assessment for high-concurrency scenarios, it is determined whether to implement intelligent control based on computing power adaptability. If so, a smart dispatch command for meal delivery is sent after control. If not, a smart dispatch command for meal delivery is sent directly, thereby improving the reliability of intelligent dispatch management for meal delivery.

[0020] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0021] This invention provides a multi-scenario intelligent dispatch and management system for meal delivery, such as... Figure 1 The diagram shows a multi-scenario intelligent dispatch and management system for food delivery. The system includes: a multi-source road condition fusion perception module, a fusion effectiveness intelligent control module, a high-concurrency computing power evaluation module, and a computing power adaptability intelligent control module.

[0022] As the first module of a multi-scenario intelligent dispatch and management system for meal delivery, the multi-source traffic condition fusion perception module is used to import multi-source heterogeneous traffic condition data from third-party map APIs and self-built monitoring systems. Based on the multi-source heterogeneous traffic condition data, the effectiveness evaluation of multi-source traffic condition data fusion is obtained, which reflects the quantitative degree of accuracy of cross-scenario traffic condition information synchronization and effectiveness of route planning data.

[0023] It should be explained that multi-source heterogeneous traffic data includes cross-regional traffic information synchronization delay, cross-source data timing alignment error, and frequency of mismatch in road network node status perception. Among them, cross-regional traffic information synchronization latency refers to the update time difference of traffic information at the same connecting node (such as intersection or road segment junction) between different scenario data sources in cross-regional routes for multi-scenario food delivery. This is achieved by extracting traffic update timestamps (such as the first record time of congestion status) of main road connecting nodes from the interface logs of third-party map APIs, and extracting traffic update timestamps of the same connecting node from the stored data of self-built monitoring systems (such as cameras and sensors in campuses / communities). The difference between the two types of timestamps is calculated to obtain the synchronization latency of each cross-regional connecting node, and then the cross-regional traffic information synchronization latency is obtained by statistical averaging. Cross-source data time-series alignment error refers to the degree of deviation between the two types of data in describing the traffic conditions of the same road segment at the same time dimension after time-series calibration of multi-source traffic data from third-party map APIs and self-built monitoring systems. This involves collecting the original timestamps and corresponding traffic data (such as congestion data) from the two types of data sources from third-party map APIs and self-built monitoring systems. The system aligns data (e.g., traffic speed, etc.) using time-series calibration algorithms such as interpolation to obtain calibrated theoretical timestamps. It then compares these theoretical timestamps with the actual occurrence times of road events (e.g., confirmed through manual annotation or high-precision monitoring backtracking) to calculate the time difference. The average of these time differences across all road segments is then used to obtain the cross-source data time-series alignment error. The road network node status perception mismatch frequency refers to the number of times within a preset unit of time that the descriptions of the road condition status (e.g., congestion / smooth traffic, traffic speed, restricted access status, etc.) of the same road network node (e.g., intersections, road segment endpoints) by the third-party map API and the self-built monitoring system are inconsistent. Real-time data collection of the road condition status of these nodes from the third-party map API and the self-built monitoring system (e.g., through API interface retrieval and monitoring data parsing) is performed. The two types of status data for the same node are compared; if the descriptions are inconsistent, it is recorded as a mismatch. The total number of mismatches within a preset unit of time is counted to obtain the road network node status perception mismatch frequency.

[0024] It should be understood that, firstly, the synchronization delay impact factor and the synchronization delay impact score are interactively processed to obtain the synchronization delay impact value. The synchronization delay impact score represents the result of the proportional analysis between the synchronization delay reference value and the synchronization delay of cross-regional road condition information. Here, the interactive processing represents a multiplication operation, and the constrained expression for the synchronization delay impact value is:

[0025] ;

[0026] In the formula, A represents the synchronization delay impact value, w1 represents the synchronization delay impact factor obtained from the intelligent scheduling management database, S0 represents the synchronization delay reference value obtained from the intelligent scheduling management database, and S represents the cross-regional road condition information synchronization delay.

[0027] Next, the alignment error impact factor and the alignment error impact score are interactively processed to obtain the alignment error impact value. The alignment error impact score represents the result of the proportion analysis between the alignment error reference value and the cross-source data time-series alignment error. The constraint expression for the alignment error impact value is as follows:

[0028] ;

[0029] In the formula, B represents the alignment error impact value, w2 represents the alignment error impact factor obtained from the intelligent scheduling management database, N0 represents the alignment error reference value obtained from the intelligent scheduling management database, and N represents the cross-source data time-series alignment error.

[0030] Secondly, the mismatch frequency impact factor and the mismatch frequency impact score are interactively processed to obtain the mismatch frequency impact value. The mismatch frequency impact score represents the result of the ratio analysis between the mismatch frequency reference value and the mismatch frequency perceived by road network nodes. The constraint expression for the mismatch frequency impact value is as follows:

[0031] ;

[0032] In the formula, C represents the mismatch frequency impact value, w3 represents the mismatch frequency impact factor obtained from the intelligent scheduling management database, F0 represents the mismatch frequency reference value obtained from the intelligent scheduling management database, and F represents the mismatch frequency perceived by the road network node status.

[0033] Finally, the impact values ​​of synchronization delay, alignment error, and mismatch frequency are superimposed to obtain an evaluation of the effectiveness of multi-source traffic data fusion. Here, superposition refers to addition, and the constraint expression for the effectiveness evaluation of multi-source traffic data fusion is:

[0034] ;

[0035] In the formula, D represents the effectiveness assessment of multi-source road condition data fusion.

[0036] It needs to be explained that the three factors—cross-regional traffic information synchronization latency, cross-source data timing alignment error, and the frequency of road network node state perception mismatch—all revolve around the effectiveness evaluation of multi-source traffic data fusion in multi-scenario intelligent scheduling management of meal delivery. They are interconnected, focusing on two core objectives: time synchronization accuracy and state perception consistency. The specific logical relationship is as follows: Cross-source data timing alignment error is a fundamental prerequisite, determining the baseline level of cross-regional traffic information synchronization latency. The essence of cross-source data timing alignment error is to measure the consistency accuracy between third-party map APIs and self-built monitoring systems in the time dimension after timing calibration—it directly reflects the effectiveness of timing calibration and is the foundation of all cross-source data collaboration. There is a positive correlation between cross-source data timing alignment error and cross-regional traffic information synchronization latency. The larger the alignment error, the longer the synchronization delay of cross-regional road condition information. The synchronization delay of cross-regional road condition information is an intermediate transmission factor that directly affects the frequency of mismatch in the state perception of road network nodes. The synchronization delay of cross-regional road condition information focuses on the information update time difference of cross-scenario connecting nodes. Its magnitude directly determines whether there is a time difference in the state description of the same road network node by the two types of data sources, thus affecting the mismatch frequency. There is a positive correlation between the synchronization delay of cross-regional road condition information and the frequency of mismatch in the state perception of road network nodes. The larger the synchronization delay of cross-regional road condition information, the higher the frequency of mismatch in the state perception of road network nodes. The frequency of mismatch in the state perception of road network nodes is the final quantification of the state consistency after the fusion of multi-source data. Its value directly reflects the comprehensive effect of cross-source temporal alignment accuracy and cross-regional information synchronization efficiency, and is the result output of the first two parameters.

[0037] Meanwhile, the synchronization delay of cross-regional traffic information, the time alignment error of cross-source data, the frequency of mismatch in road network node status perception, and the effectiveness evaluation of multi-source traffic data fusion all show a negative correlation. That is, the smaller the value of the first three, the higher the effectiveness of multi-source traffic data fusion; conversely, the larger the value, the lower the fusion effectiveness. The specific positive and negative correlation logic and details are as follows: There is a negative correlation between the time alignment error of cross-source data and the effectiveness evaluation of multi-source traffic data fusion. The larger the time alignment error of cross-source data, the more significant the time scale misalignment between the two types of data, namely, the third-party map API and the self-built monitoring system. Even if data integration is carried out subsequently, the inconsistent time base will lead to spatiotemporal mismatch in traffic information, resulting in a lower effectiveness evaluation of multi-source traffic data fusion. There is also a negative correlation between the synchronization delay of cross-regional traffic information and the effectiveness evaluation of multi-source traffic data fusion. The longer the synchronization delay of traffic information, the more likely it is that the main road data will show smooth traffic while the internal branch roads are actually congested. In this case, the fused data cannot accurately reflect the actual traffic conditions of cross-regional roads, and route planning based on this data will frequently fail (such as guiding riders into congested sections), directly reducing the support of data fusion for scheduling decisions and resulting in a lower effectiveness assessment of multi-source traffic data fusion. There is a negative correlation between the frequency of mismatch in road network node status perception and the effectiveness assessment of multi-source traffic data fusion. The higher the frequency of mismatch in road network node status perception, the more contradictory information there is in the fused data. The system cannot accurately judge the true status of road network nodes, which will lead to a surge in route planning error rate and increased extra time for riders. Multi-source data fusion will not only fail to improve scheduling efficiency, but will also mislead decision-making, resulting in a lower effectiveness assessment of multi-source traffic data fusion.

[0038] As the second module of a multi-scenario intelligent dispatch and management system for meal delivery, the fusion effectiveness intelligent control module is used to determine whether to perform fusion effectiveness intelligent control based on the fusion effectiveness assessment of multi-source road condition data. If yes, it generates updated fusion road condition information after control and transmits it to the high-concurrency computing power assessment module. If no, it directly generates updated fusion road condition information and transmits it to the high-concurrency computing power assessment module. Fusion effectiveness intelligent control includes time granularity control and cross-validation node number control.

[0039] Furthermore, the specific steps for determining whether to implement intelligent control of fusion effectiveness are as follows: if the multi-source traffic data fusion effectiveness assessment is higher than or equal to the effectiveness setting value, then intelligent control of fusion effectiveness is not implemented; if the multi-source traffic data fusion effectiveness assessment is lower than the effectiveness setting value, then it is determined whether to implement time granularity control based on the heterogeneity frequency of multi-source traffic data. If so, then it is determined whether to implement cross-validation node number control after control; if not, then it is determined directly whether to implement cross-validation node number control.

[0040] It's important to explain that in intelligent scheduling and management of food delivery across multiple scenarios, including campuses, commercial districts, and communities, the heterogeneous frequency of multi-source road condition data (i.e., the difference in update frequency between real-time road conditions on main roads in commercial districts provided by third-party map APIs and the traffic status of internal campus roads and community alleys captured by self-built monitoring) directly determines the difficulty of matching these two types of data in the time dimension. The core function of dynamic interpolation time granularity is to bridge this frequency difference by generating pseudo-real-time data points. Therefore, dynamically adjusting the granularity based on heterogeneous frequencies has a clear logical necessity. This dynamic adjustment mechanism essentially achieves a balance between "time alignment accuracy" and "computing efficiency" by adapting the interpolation granularity to the data frequency differences, ultimately improving the effectiveness of multi-source road condition data fusion. This provides a more accurate temporal basis for cross-regional route planning in multi-scenario food delivery (such as from restaurants in commercial districts to campus dormitories, or from community convenience stores to residential buildings), thereby shortening delivery time and reducing the risk of order delays.

[0041] like Figure 2 The diagram shows a flowchart of the cross-validation node quantity control process in a multi-scenario intelligent scheduling and management system for meal delivery provided by an embodiment of the present invention. The specific logic is as follows: First, determine the data cache preloading time window. If the data cache preloading time window is greater than the upper limit of the time window reference, then the result of the arithmetic average of the time window correction and the validity correction, rounded up, is input into the time window mapping table for index lookup to obtain the reduction in the number of validation nodes. The current number of dual-source cross-validation nodes is then reduced by the reduction in the number of validation nodes to obtain the adjusted number of dual-source cross-validation nodes. If the data cache preloading time window is within the time window reference interval, then no cross-validation node quantity control is performed. If the data cache preloading time window is less than the lower limit of the time window reference, then the result of the arithmetic average of the time window correction and the validity correction, rounded down, is input into the time window mapping table for index lookup to obtain the increase in the number of validation nodes. The current number of dual-source cross-validation nodes is then coupled with the increase in the number of validation nodes to obtain the adjusted number of dual-source cross-validation nodes.

[0042] As further clarification, the specific steps for determining whether to perform time-granularity control based on the heterogeneous frequency of multi-source road condition data are as follows:

[0043] If the heterogeneous frequency of multi-source road condition data is higher than the upper limit of the heterogeneous frequency reference, the heterogeneous frequency correction amount is input into the granularity mapping table for index lookup to obtain the granularity adjustment factor. It is then determined whether the granularity adjustment factor is greater than the granularity adjustment reference value. This pre-judgment process can accurately locate the time dimension matching problem of multi-source data, providing a basis for subsequent targeted regulation and ensuring the basic quality of road condition data fusion from the source. The heterogeneous frequency correction amount represents the positive difference between the heterogeneous frequency of multi-source road condition data and the upper limit of the heterogeneous frequency reference.

[0044] If so, the result of harmonic averaging of the granularity adjustment correction and the effectiveness correction is input into the granularity mapping table for index lookup to obtain the granularity reduction amount. The difference between the current dynamic interpolation time granularity and the granularity reduction amount is processed to obtain the adjusted dynamic interpolation time granularity. By refining the granularity and increasing pseudo-real-time data points, the data gaps caused by high-frequency differences can be effectively bridged, significantly improving the accuracy of cross-scenario road condition information synchronization and providing more accurate time-series data support for subsequent delivery route planning. The granularity adjustment correction amount represents the positive difference between the granularity adjustment factor and the granularity adjustment reference value, while the effectiveness correction amount represents the negative difference between the effectiveness assessment of multi-source road condition data fusion and the effectiveness setting value.

[0045] If not, the result of harmonic averaging of the granularity adjustment reference and the validity correction is input into the granularity mapping table for index lookup to obtain the granularity increase. The current dynamic interpolation time granularity is coupled with the granularity increase to obtain the adjusted dynamic interpolation time granularity. By appropriately coarsening the granularity, unnecessary computing power consumption can be reduced and resource redundancy can be avoided while ensuring the basic validity of the data. Sufficient computing power is reserved for core scheduling tasks in high-concurrency scenarios, ensuring the efficient generation and issuance of intelligent scheduling instructions for meal delivery. The granularity adjustment reference is used to reflect the negative difference between the granularity adjustment factor and the granularity adjustment reference value.

[0046] In this embodiment, the judgment process for time granularity control achieves dual technical value through a closed-loop operation of "frequency threshold determination - adjustment factor verification - precise quantitative de-adjustment": On the one hand, by fine-tuning for scenarios where heterogeneous frequencies exceed the upper limit, the granularity is refined to supplement pseudo-real-time data points when the adjustment factor meets the standard, effectively compensating for the frequency difference gaps in multi-source road condition data, significantly improving the accuracy of cross-scenario road condition information synchronization, and providing high-precision time-series data support for multi-scenario meal delivery route planning; on the other hand, when the adjustment factor does not meet the standard, the granularity is appropriately coarsened, minimizing ineffective computing power consumption and avoiding resource redundancy while ensuring the basic validity of the data, reserving sufficient computing power for high-concurrency scenarios such as peak order periods, ensuring low-latency generation and issuance of core scheduling instructions, and ultimately achieving synergistic optimization of the multi-source road condition data fusion quality and high-concurrency scenario scheduling efficiency, providing key technical guarantees for the stable and efficient operation of the intelligent delivery system.

[0047] As a further specific explanation, determining whether to perform time-granularity control also includes:

[0048] If the heterogeneous frequencies of multi-source road condition data are within the heterogeneous frequency reference range, the heterogeneous frequencies are input into the granularity mapping table for index lookup to obtain the granularity control factor. It is then determined whether the granularity control factor is greater than the granularity control reference value. This precise judgment logic ensures that control is only initiated when the data frequency is within a reasonable fluctuation range, avoiding invalid operations and ensuring the efficient use of system resources. The heterogeneous frequency reference range represents the closed interval formed by the lower limit and upper limit of the heterogeneous frequency reference.

[0049] If so, the result of harmonic averaging of the granularity control reference value and the effectiveness correction value is input into the granularity mapping table for index lookup to obtain the granularity adjustment amount. The current dynamic interpolation time granularity is coupled with the granularity adjustment amount to obtain the adjusted dynamic interpolation time granularity. The moderately coarsened granularity can reduce the redundant generation of pseudo-real-time data points and reduce computing power consumption while meeting the basic synchronization requirements of road condition data. This allows for more computing resources to be reserved for core delivery scheduling tasks in high-concurrency scenarios, ensuring a rapid response to scheduling instructions. The granularity control reference value represents the positive difference between the granularity control factor and the granularity control reference value.

[0050] If not, the result of harmonic averaging of the granularity adjustment correction and the effectiveness correction is input into the granularity mapping table for index lookup to obtain the granularity reduction amount. The current dynamic interpolation time granularity is then interpolated with the granularity reduction amount to obtain the adjusted dynamic interpolation time granularity. The refined granularity can further improve the time alignment accuracy of multi-source data, effectively compensate for information deviations caused by subtle frequency differences, and ensure that the fused traffic data is more in line with the actual traffic conditions in multiple scenarios such as campuses, business districts, and communities. This provides more accurate time-series support for delivery route planning and reduces the problem of route detours or timeouts caused by data errors. The granularity adjustment correction amount represents the negative difference between the granularity adjustment factor and the granularity adjustment reference value.

[0051] If the heterogeneity frequency of multi-source traffic data is lower than the reference lower limit of heterogeneity frequency, time granularity control will not be performed. This will maintain the current stable granularity parameters, ensuring the consistency of traffic data fusion and avoiding system overhead caused by excessive control, thus achieving a dynamic balance between data processing efficiency and fusion quality.

[0052] In this embodiment, the time-granularity control logic achieves multiple technical optimizations through precise application of policies across different scenarios: First, for heterogeneous frequencies within the reference range, the pre-processing logic of "frequency query - factor determination" ensures that control is only initiated when data fluctuations are reasonable, avoiding ineffective operations and saving system resources; when the control factor exceeds the reference value, the granularity is appropriately coarsened by increasing it, reducing the generation of redundant pseudo-data while meeting the basic synchronization requirements of road condition data, reducing computing power consumption, reserving sufficient computing resources for high-concurrency delivery scheduling, and ensuring rapid response to core instructions; when the control factor does not reach the reference value... By refining the granularity of the processing, the time alignment accuracy of multi-source data is improved, compensating for information deviations caused by subtle frequency differences. This makes the fused traffic data more closely match the actual traffic conditions in multiple scenarios, providing precise timing support for route planning and reducing delivery delays and detours. For scenarios with heterogeneous frequencies below the reference lower limit, no adjustments are performed to maintain granularity stability. This ensures data fusion consistency while avoiding the system overhead of excessive adjustments. Ultimately, this achieves synergistic optimization of multi-source traffic data fusion quality, computing power utilization efficiency, and multi-scenario delivery scheduling accuracy, providing technical support for the efficient and stable operation of the intelligent delivery system.

[0053] It should be understood that in the intelligent scheduling and management of multi-scenario meal delivery involving campuses, business districts, and communities, the strategy of dynamically adjusting the number of dual-source cross-validation nodes based on the data cache preloading time window has strong practical value: the former (data cache preloading time window) can optimize the synchronization efficiency of cross-regional delivery data by caching data such as road conditions and capacity distribution of cross-scenario connecting nodes (such as the intersection of main roads in business districts and communities, and the area around campus takeout pick-up points) in advance, avoiding scheduling decision delays caused by real-time data transmission delays; the latter (number of dual-source cross-validation nodes) can improve the consistency of capacity status perception in different scenarios by increasing or decreasing the cross-verification nodes of rider location data and delivery station monitoring data, reducing problems such as order re-dispatch and timeouts caused by errors in a single data source. The two have a logical basis for achieving collaborative control through scenario association. The size of the data cache preloading time window directly reflects the system's depth of prediction of cross-regional node traffic data and its reliance on historical data. For example, during the lunch rush hour in a commercial district, increasing the preloading window allows for caching congestion trend data for a longer period. At this time, the need for real-time verification is reduced, and the number of cross-verification nodes can be appropriately reduced, saving computing power and ensuring scheduling response speed. However, during off-peak hours in a community, decreasing the preloading window results in insufficient historical data support, requiring an increase in the number of verification nodes to ensure the accuracy of capacity status judgment. The core of this dynamic adjustment mechanism is to utilize the complementary relationship between "historical cache completeness" and "real-time verification strength" to ultimately achieve synergistic optimization of cross-regional data synchronization efficiency and status perception consistency in multi-scenario meal delivery. This improves the overall effectiveness of multi-source traffic and capacity data fusion, providing accurate and efficient data support for intelligent scheduling decisions in different scenarios, thereby reducing delivery delay rates and improving the user dining experience.

[0054] As a further explanation, the specific steps for determining whether to perform cross-validation node number adjustment are as follows:

[0055] If the data cache preloading time window is greater than the time window reference upper limit, the result of the arithmetic average of the time window correction and the validity correction is rounded up and input into the time window mapping table for index lookup to obtain the reduction in the number of verification nodes. The current number of dual-source cross-validation nodes is reduced by the reduction in the number of verification nodes to obtain the adjusted number of dual-source cross-validation nodes. This can reduce redundant verification operations and reduce computing power consumption while ensuring basic consistency of state awareness, freeing up more computing resources for core delivery scheduling tasks in high-concurrency scenarios and improving the efficiency of scheduling instruction processing. The time window correction represents the positive difference between the data cache preloading time window and the time window reference upper limit.

[0056] If the data cache preloading time window is within the time window reference interval, then the cross-validation node quantity adjustment will not be performed. This indicates that the current cached data volume and real-time verification requirements have reached an optimal balance, which can maintain the stability and accuracy of multi-source road condition data fusion, and avoid unnecessary adjustment operations that cause system overhead, thus ensuring the efficient advancement of intelligent delivery scheduling. The time window reference interval represents the closed interval formed by the lower limit of the time window reference and the upper limit of the time window reference.

[0057] If the data cache preloading time window is less than the lower limit of the time window reference, the result of the arithmetic average of the time window comparison and the validity correction is rounded down and input into the time window mapping table for index lookup to obtain the increase in the number of verification nodes. The current number of dual-source cross-validation nodes is coupled with the increase in the number of verification nodes to obtain the adjusted number of dual-source cross-validation nodes. The increased number of verification nodes can improve the consistency of rider and unmanned vehicle capacity status perception, reduce scheduling deviations caused by errors in a single data source, and ensure that the fused road condition and capacity data are more in line with the delivery needs of multiple scenarios such as campuses, business districts, and communities. This provides accurate data support for route planning and order allocation, reduces the risk of delivery delays, and the time window comparison represents the negative difference between the data cache preloading time window and the upper limit of the time window reference.

[0058] In this embodiment, the dual-source cross-validation node quantity control mechanism achieves multi-dimensional technical optimization by precisely implementing policies based on different states of the data cache preloading time window: when the window is larger than the reference upper limit, the number of validation nodes is reduced to reduce redundant operations and lower computing power consumption while ensuring basic consistency of state perception, thereby freeing up computing resources for core delivery scheduling in high-concurrency scenarios and improving instruction processing efficiency; when the window is within the reference range, no control is performed to maintain the optimal balance between cached data and real-time validation, ensuring the stability and accuracy of multi-source road condition data fusion while avoiding unnecessary system overhead, thus facilitating the efficient advancement of intelligent delivery scheduling; when the window is smaller than the reference lower limit, the number of validation nodes is increased to strengthen the consistency of capacity state perception, reduce scheduling deviations caused by errors in a single data source, ensure that the fused data meets the delivery needs of multiple scenarios, provide accurate support for route planning and order allocation, effectively reduce the risk of delivery delays, and ultimately achieve a synergistic improvement in data fusion quality, computing power utilization efficiency, and delivery scheduling accuracy, providing a strong guarantee for the stable and efficient operation of the multi-scenario intelligent meal delivery scheduling system.

[0059] As the third module of a multi-scenario intelligent scheduling and management system for food delivery, the high-concurrency computing power assessment module is used to obtain high-concurrency computing power data under high-concurrency scenarios. Based on the high-concurrency computing power data, a high-concurrency scenario computing power adaptability assessment is obtained to reflect the quantitative degree of adaptability between delivery timeliness and intelligent scheduling of food delivery under high-concurrency scenarios.

[0060] It should be noted that high-concurrency computing power data includes the effectiveness assessment of multi-source road condition data fusion, concurrent scheduling instruction processing throughput, and cross-capacity scheduling collaboration latency. Among them, concurrent scheduling instruction processing throughput refers to the total amount of valid scheduling instructions that can be successfully generated, verified, and issued to riders or unmanned delivery vehicles within a preset unit of time in high-concurrency scenarios such as peak order periods (e.g., lunch and dinner peaks). Valid scheduling instructions refer to those that include information such as complete route planning, task priority, and time constraints. From the core processing module logs, extract the number of successfully generated and marked as valid scheduling instructions per second during high-concurrency periods. Filter out invalid instructions caused by missing data (e.g., unsynchronized road condition information) or calculation errors (e.g., path planning logic conflicts). Only count instructions that can be received and executed by riders / autonomous vehicle terminals. Accumulate the number of valid instructions within each minimum preset time window to obtain the concurrent scheduling instruction processing throughput. Cross-capacity scheduling collaboration latency (referring to the time required for collaborative scheduling of multiple types of capacity (e.g., riders, autonomous delivery vehicles, smart pickup lockers) in high-concurrency scenarios (e.g., riders needing to reassign some orders to idle autonomous vehicles due to high load, or autonomous vehicles needing to relay deliveries with riders at community entrances) is calculated from the triggering of the collaboration requirement to the generation and issuance of the collaboration protocol. The time difference between the same dispatch instruction and the time difference between the two is calculated. The timestamps of the cross-capacity collaborative events triggered by the cross-capacity collaborative events are extracted from the logs of the capacity monitoring module. The timestamps of the cross-capacity collaborative events include the first detection time of scenarios such as rider overload, insufficient battery power of unmanned vehicles, and overflow of food collection lockers requiring manual assistance. The timestamps of the instruction generation (i.e., the final confirmation time of the instruction containing complete reassignment rules, handover points, and timeliness requirements) and the timestamps of the instruction issuance (the time when the instruction is transmitted to the target capacity terminal) are extracted from the logs of the dispatch decision module. The time difference between the two is calculated from the "time of triggering the collaborative requirement" to the "time of issuing the collaborative instruction". This is the single cross-capacity dispatch collaborative delay. The cross-capacity dispatch collaborative delay is obtained by statistically analyzing all collaborative events during high-concurrency periods and taking the average value.

[0061] It needs to be explained that, firstly, the fusion effectiveness impact factor and the effectiveness assessment impact score are interactively processed to obtain the effectiveness assessment impact value. The effectiveness assessment score represents the result of the proportional analysis between the multi-source road condition data fusion effectiveness assessment and the fusion effectiveness reference value. The constraint expression for the effectiveness assessment impact value is as follows:

[0062] ;

[0063] In the formula, P represents the effectiveness assessment impact value, r1 represents the fusion effectiveness impact factor obtained from the intelligent dispatch management database, Q0 represents the fusion effectiveness reference value obtained from the intelligent dispatch management database, and Q represents the multi-source traffic data fusion effectiveness assessment.

[0064] Next, the throughput impact factor and throughput impact score are interactively processed to obtain the throughput impact value. The throughput impact score represents the result of the ratio analysis between the concurrent scheduling instruction processing throughput and the throughput reference value. The constraint expression for the throughput impact value is as follows:

[0065] ;

[0066] In the formula, M represents the throughput impact value, r2 represents the throughput impact factor obtained from the intelligent scheduling management database, Y0 represents the throughput reference value obtained from the intelligent scheduling management database, and Y represents the throughput of concurrent scheduling instruction processing.

[0067] Secondly, the collaborative delay impact factor and the collaborative delay impact score are interactively processed to obtain the collaborative delay impact value. The collaborative delay impact score represents the result of the proportional analysis between the collaborative delay reference value and the cross-capacity scheduling collaborative delay. The constraint expression for the collaborative delay impact value is as follows:

[0068] ;

[0069] In the formula, K represents the collaborative delay impact value, r3 represents the collaborative delay impact factor obtained from the intelligent scheduling management database, T0 represents the collaborative delay reference value obtained from the intelligent scheduling management database, and T represents the cross-capacity scheduling collaborative delay.

[0070] Finally, the impact values ​​of effectiveness assessment, throughput, and collaborative latency are summed to obtain the computing power adaptability assessment for high-concurrency scenarios. The constraint expression for the computing power adaptability assessment for high-concurrency scenarios is as follows:

[0071] ;

[0072] In the formula, U represents the computing power adaptability assessment for high-concurrency scenarios.

[0073] It's important to clarify that the effectiveness assessment of multi-source traffic data fusion, the throughput of concurrent dispatch instructions, and the delay of cross-capacity dispatch coordination all revolve around the core objective of intelligent dispatch management for multi-scenario food delivery. These three aspects are interconnected and influence each other, with the specific logical relationship as follows: The effectiveness assessment of multi-source traffic data fusion is a fundamental prerequisite. The throughput of concurrent dispatch instructions measures the number of valid instructions processed within a preset unit of time. Valid instructions are based on accurate traffic data. If the effectiveness assessment of multi-source traffic data fusion is low, the generated dispatch instructions may contain planning errors (such as guiding riders into actual congested sections). Even if these invalid instructions are processed quickly, they require recalculation, which consumes additional computing power. Resource constraints lead to a decrease in the throughput of concurrent scheduling instructions. Cross-capacity collaborative scheduling (such as rider and unmanned vehicle relay delivery) requires precise planning of handover points and routes based on road condition data. If there are deviations in the fusion of multi-source road condition data, the lower the effectiveness assessment of multi-source road condition data fusion, the longer the delay in cross-capacity collaborative scheduling will be due to the suspension of calculations and re-verification of data during the generation of collaborative instructions. The throughput of concurrent scheduling instructions is the core support and affects the efficiency limit of cross-capacity collaborative scheduling delay. If the throughput of concurrent scheduling instructions is low, regular scheduling instructions will occupy most of the computing power during high-concurrency periods, and cross-capacity collaborative instructions will have to wait for computing power to be idle before they can start calculations, resulting in a longer delay in cross-capacity collaborative scheduling.

[0074] Meanwhile, the correlation between the effectiveness assessment of multi-source traffic data fusion, the throughput of concurrent scheduling command processing, the cross-capacity scheduling coordination delay, and the computing power adaptability assessment in high-concurrency scenarios can be divided into two categories: positive and negative correlations. The specific logic and details of the positive and negative correlations are as follows: There is a positive correlation between the effectiveness assessment of multi-source traffic data fusion and the computing power adaptability assessment in high-concurrency scenarios. The higher the effectiveness assessment of multi-source traffic data fusion, the more accurate and timely the traffic information is. When the system generates scheduling commands, it does not need to repeatedly verify and correct the data, which can reduce invalid calculations and allow computing power to be concentrated on serving effective scheduling needs. The higher the computing power adaptability assessment in high-concurrency scenarios, the better. High; there is a positive correlation between the throughput of concurrent scheduling instructions and the computing power adaptability assessment in high-concurrency scenarios. The higher the throughput of concurrent scheduling instructions, the more sufficient the computing power is, and there is no need to queue instructions, thus avoiding scheduling delays caused by computing power bottlenecks. The higher the computing power adaptability assessment in high-concurrency scenarios, the better. There is a negative correlation between the cross-capacity scheduling collaboration delay and the computing power adaptability assessment in high-concurrency scenarios. The longer the cross-capacity scheduling collaboration delay, the more it means that during high-concurrency periods, regular scheduling instructions have already occupied most of the computing power, and more complex instructions such as cross-capacity collaboration, which consume more computing power, need to queue, resulting in response delays. The lower the computing power adaptability assessment in high-concurrency scenarios, the worse.

[0075] As the fourth module of a multi-scenario intelligent scheduling and management system for meal delivery, the intelligent computing power adaptability control module is used to determine whether to perform intelligent computing power adaptability control based on the computing power adaptability assessment of high-concurrency scenarios. If yes, it sends a meal delivery intelligent scheduling instruction after the control; otherwise, it sends a meal delivery intelligent scheduling instruction directly. The intelligent computing power adaptability control includes the threshold control of the length of the low-priority instruction cache queue and the control of the granularity of collaborative task splitting.

[0076] Furthermore, the specific steps for determining whether to execute intelligent control of computing power adaptability are as follows: if the computing power adaptability assessment for high-concurrency scenarios is greater than or equal to the adaptability setting value, then intelligent control of computing power adaptability will not be executed; if the computing power adaptability assessment for high-concurrency scenarios is less than the adaptability setting value, then it is determined whether to execute the threshold control of the low-priority instruction cache queue length based on the current preset number of meal delivery orders per unit time. If yes, then it is determined whether to execute the granular control of collaborative task splitting after the control; if no, then it is directly determined whether to execute the granular control of collaborative task splitting.

[0077] It should be understood that in the intelligent scheduling and management of meal delivery in various scenarios such as campuses, business districts, and communities, the current preset order volume per unit time directly reflects the system's workload intensity. For example, during peak hours in business districts, when orders surge, or during breaks in schools, the system must prioritize the efficient processing of core scheduling instructions such as real-time route planning and cross-capacity relay reassignment. During off-peak hours, when community orders are mainly scattered, the system load is relatively relaxed, allowing for the timely processing of lower-priority instructions. The core function of the low-priority instruction cache queue length threshold is to balance the processing efficiency of high and low-priority tasks through dynamic allocation of computing resources. Therefore, adjusting this threshold based on order volume has a clear logical necessity. This dynamic adjustment mechanism essentially achieves a dynamic balance between "core scheduling efficiency" and "full task integrity" by matching the cache threshold with the real-time business load. This ensures that the intelligent scheduling system for meal delivery in various scenarios can operate efficiently under different order volume scenarios, meeting the immediate delivery needs of massive orders during peak hours while ensuring the stability and orderliness of the entire system process during off-peak hours.

[0078] As further explained in detail, the specific steps for adjusting the threshold length of the low-priority instruction cache queue are as follows:

[0079] If the number of meal delivery orders within the current preset unit time exceeds the upper limit of the order volume setting, the concurrency pressure value is input into the cache threshold mapping table for index lookup to obtain the low-priority instruction cache queue length threshold adjustment factor. It is then determined whether the low-priority instruction cache queue length threshold adjustment factor is greater than the threshold adjustment reference value. This pre-judgment process can accurately identify the computing power allocation requirements in high-concurrency scenarios, providing a scientific basis for subsequent threshold control and avoiding resource mismatch caused by blind adjustment. The concurrency pressure value represents the positive difference between the number of meal delivery orders within the current preset unit time and the upper limit of the order volume setting.

[0080] If so, the result of harmonic averaging of the threshold adjustment correction amount and the computing power adaptation correction amount is input into the cache threshold mapping table for index lookup to obtain the threshold increase amount. The current low-priority instruction cache queue length threshold is superimposed with the threshold increase amount to obtain the adjusted low-priority instruction cache queue length threshold. After the threshold is increased, more low-priority instructions such as rider status periodic updates and historical data statistics can be temporarily cached, and computing power can be prioritized for core scheduling tasks such as route planning and order reassignment. This effectively avoids computing power contention in high-concurrency scenarios, ensures low-latency processing of core instructions, and ensures the timeliness of delivery scheduling. The threshold adjustment correction amount represents the positive difference between the low-priority instruction cache queue length threshold adjustment factor and the threshold adjustment reference value, while the computing power adaptation correction amount represents the negative difference between the computing power adaptation assessment and the adaptation setting value in high-concurrency scenarios.

[0081] If not, the harmonic average of the threshold adjustment reference and the computing power adaptation correction is input into the cache threshold mapping table for index lookup to obtain the threshold reduction amount. The difference between the current low-priority instruction cache queue length threshold and the threshold reduction amount is processed to obtain the adjusted low-priority instruction cache queue length threshold. The moderately reduced threshold can avoid information lag caused by excessive caching of low-priority instructions. While ensuring the core scheduling computing power, it ensures the orderly progress of auxiliary tasks such as rider status synchronization and data archiving, maintains the stable operation of the entire system process, and achieves a dynamic balance between core scheduling efficiency and the integrity of all tasks in high-concurrency scenarios. The threshold adjustment reference amount represents the negative difference between the low-priority instruction cache queue length threshold adjustment factor and the threshold adjustment reference value.

[0082] In this embodiment, the low-priority instruction cache queue length threshold control mechanism achieves a dual improvement in optimized allocation of computing resources and system operating efficiency through accurate judgment and dynamic adaptation in high-concurrency scenarios: First, by using the pre-process of "concurrency pressure value query - adjustment factor verification", the computing power allocation demand when the order volume exceeds the set upper limit is accurately identified, avoiding resource mismatch caused by blind adjustment, and laying the foundation for scientific scheduling; when the adjustment factor is greater than the reference value, more low-priority instructions are temporarily cached by raising the threshold, and computing power is released to core scheduling tasks first, effectively avoiding computing power contention, ensuring low-latency processing of key instructions such as route planning and order reassignment, and ensuring the timeliness of delivery scheduling in high-concurrency scenarios; when the adjustment factor does not reach the reference value, the threshold is appropriately lowered, ensuring core computing power while avoiding information lag caused by excessive caching of low-priority instructions, ensuring the orderly progress of auxiliary tasks such as rider status synchronization and data archiving, achieving a dynamic balance between core scheduling efficiency and the integrity of all tasks, and ultimately providing strong support for the stable and efficient operation of the multi-scenario intelligent scheduling system for meal delivery during peak order periods.

[0083] As a further explanation, the implementation of low-priority instruction cache queue length threshold control also includes:

[0084] If the number of meal delivery orders within the current preset unit time is within the order quantity setting range, the low-priority instruction cache queue length threshold adjustment will not be executed. This maintains the rationality of the existing computing power allocation, ensures the orderly progress of core scheduling tasks and low-priority tasks, and avoids unnecessary adjustment operations that consume system resources, thus ensuring the efficient and stable operation of meal delivery scheduling in multiple scenarios. The order quantity setting range represents the closed interval formed by the lower bound and the upper bound of the order quantity setting.

[0085] If the number of meal delivery orders within the current preset unit time is less than the lower bound of the order quantity setting, the concurrency relaxation is input into the cache threshold mapping table for index lookup to obtain the low-priority instruction cache queue length threshold adjustment factor. It is then determined whether the low-priority instruction cache queue length threshold adjustment factor is greater than the threshold adjustment reference value. This precise determination can adjust the threshold strategy in a targeted manner according to the computing power redundancy situation in low-concurrency scenarios, thereby achieving refined utilization of computing power resources. The concurrency relaxation represents the negative difference between the number of meal delivery orders within the current preset unit time and the lower bound of the order quantity setting.

[0086] If so, the result of harmonic averaging of the threshold control reference value and the computing power adaptation correction value is input into the cache threshold mapping table for index lookup to obtain the threshold reduction amount. The difference between the current low-priority instruction cache queue length threshold and the threshold reduction amount is processed to obtain the adjusted low-priority instruction cache queue length threshold. After the threshold is reduced, the cache restriction of low-priority instructions can be reduced, allowing tasks such as rider status updates and historical data statistics to receive timely computing power support, avoiding information accumulation, ensuring the real-time and integrity of system data, and making full use of redundant computing power to improve the overall processing efficiency of the system. The threshold control reference value represents the positive difference between the low-priority instruction cache queue length threshold control factor and the threshold control reference value.

[0087] If not, the result of harmonic averaging of the threshold adjustment correction and the computing power adaptation correction is input into the cache threshold mapping table for index lookup to obtain the threshold increase. The current low-priority instruction cache queue length threshold is superimposed with the threshold increase to obtain the adjusted low-priority instruction cache queue length threshold. The moderately increased threshold can reserve a certain amount of cache space in low-concurrency scenarios, avoid computing power allocation chaos when the order volume suddenly increases, ensure that the system has the ability to respond quickly to business fluctuations, and achieve a dynamic balance between resource reservation and current task processing in low-concurrency state. The threshold adjustment correction represents the negative difference between the low-priority instruction cache queue length threshold adjustment factor and the threshold adjustment reference value.

[0088] In this embodiment, the low-priority instruction cache queue length threshold control mechanism achieves refined management of computing resources and optimization of system performance for different order volume scenarios: when the order volume is within a set range, no control is performed to maintain the rationality of the existing computing power allocation, ensuring the orderly progress of core scheduling and low-priority tasks, while avoiding ineffective operations that consume resources, thus ensuring efficient and stable delivery scheduling; when the order volume is below a set lower bound, the strategy is adjusted in a targeted manner based on the computing power redundancy situation in low-concurrency scenarios through precise judgment of "concurrency relaxation query - control factor verification": if the control factor exceeds the reference value, lowering the threshold can reduce the low-priority instruction cache restriction, allowing auxiliary tasks to obtain computing power support in a timely manner, avoiding information accumulation and making full use of redundant computing power; if the control factor does not reach the reference value, appropriately raising the threshold can reserve cache space, ensuring that the system can quickly respond to subsequent sudden increases in order volume, achieving a dynamic balance between resource reservation and current task processing, and ultimately comprehensively improving the adaptability and operational efficiency of the multi-scenario intelligent scheduling system for meal delivery under different business loads.

[0089] It should be understood that in intelligent scheduling and management of multi-scenario food delivery (covering differentiated scenarios such as campuses, business districts, and communities, and needing to cope with complex demands such as order surges during lunch and evening peak hours and cross-scenario delivery relay), the delay in cross-capacity scheduling directly reflects the efficiency of cross-capacity collaborative instructions (such as order relay between riders in business districts and unmanned vehicles in communities, and task reassignment by different delivery teams within campuses) from triggering to execution. The granularity of collaborative task decomposition determines the balance between the parallel processing capability and communication cost of collaborative tasks. Therefore, dynamically adjusting the latter based on the former has a clear logical necessity. This dynamic adjustment mechanism essentially adapts the task decomposition strategy to the real-time collaborative efficiency requirements, maximizing computing power utilization efficiency while ensuring the timeliness of cross-capacity scheduling in multi-scenario food delivery (such as avoiding order delays and meeting users' immediate dining needs). It avoids collaborative delay fluctuations caused by rigid granularity (such as order backlogs caused by excessive delays in community deliveries) or resource waste (such as idle computing power during off-peak hours on campuses), providing precise support for efficient delivery scheduling in multiple scenarios.

[0090] like Figure 3 This is a flowchart illustrating the collaborative task splitting granularity control process of a multi-scenario intelligent dispatch management system for meal delivery provided by an embodiment of the present invention. The specific logic is as follows: First, determine the cross-capacity scheduling collaborative delay. If the cross-capacity scheduling collaborative delay is within the collaborative delay threshold range, then no collaborative task splitting granularity control is performed. If the cross-capacity scheduling collaborative delay is greater than the upper limit of the collaborative delay, then the result of harmonic averaging the collaborative delay correction and the computing power adaptation correction is input into the granularity adjustment mapping table to obtain the subtask splitting quantity adjustment coefficient. Then, determine whether the subtask splitting quantity adjustment coefficient is greater than the splitting quantity adjustment set value. If yes, then the result of rounding up the splitting quantity adjustment offset and the current collaborative task splitting granularity is used as the adjusted collaborative task splitting granularity. If no... The lowered collaborative task splitting granularity is obtained by interacting the splitting quantity adjustment reference amount with the current collaborative task splitting granularity and rounding it down. If the cross-capacity scheduling collaborative delay is less than the lower bound of the collaborative delay, the result of harmonic averaging the collaborative delay reference amount and the computing power adaptation correction amount is input into the granularity adjustment mapping table to obtain the subtask splitting quantity control factor. It is then determined whether the subtask splitting quantity control factor is greater than the splitting quantity control setting value. If so, the lowered collaborative task splitting granularity is obtained by interacting the splitting quantity control compensation amount with the current collaborative task splitting granularity and rounding it down. If not, the uppered collaborative task splitting granularity is obtained by interacting the splitting quantity control reference amount with the current collaborative task splitting granularity and rounding it up.

[0091] As a further explanation, the specific process for determining whether to perform collaborative task splitting granularity control is as follows:

[0092] If the cross-capacity scheduling coordination delay is within the coordination delay threshold range, the granular control of coordination task splitting will not be executed. This setting can avoid unnecessary control operations that would waste system resources, while maintaining the stability of coordination task processing, ensuring the smoothness of cross-capacity cooperation in multi-scenario meal delivery, and ensuring that the order delivery rhythm is not disturbed. The coordination delay threshold range represents the closed interval formed by the lower bound and the upper bound of the coordination delay.

[0093] If the cross-capacity scheduling coordination delay exceeds the upper limit of the coordination delay, the harmonic averaging result of the coordination delay correction and the computing power adaptation correction is input into the granularity adjustment mapping table to obtain the subtask splitting quantity adjustment coefficient. It is then determined whether the subtask splitting quantity adjustment coefficient is greater than the splitting quantity adjustment setting value. This precise judgment process, combined with the delay status and actual computing power adaptation, allows for the development of targeted splitting strategies, avoiding coordination chaos caused by blind adjustments. If so, the rounded result of the interaction between the splitting quantity adjustment offset and the current coordination task splitting granularity is used as the adjusted coordination task splitting granularity. More refined task splitting enables parallel and efficient processing of subtasks, effectively shortening the processing time of cross-capacity coordination instructions, alleviating scheduling delay problems, and ensuring that orders are processed efficiently as planned. Delivery; if not, the rounded result of the interaction between the split quantity adjustment reference and the current collaborative task split granularity is used as the adjusted collaborative task split granularity. Appropriately coarsening the split granularity can reduce the communication overhead and coordination cost between subtasks, while avoiding excessive dispersion of computing power, ensuring the processing quality of core collaborative tasks, achieving a dynamic balance between cross-capacity scheduling efficiency and computing power utilization efficiency, and ensuring the stability and timeliness of cross-capacity collaborative scheduling in high-concurrency scenarios. The collaborative delay correction amount represents the positive difference between the cross-capacity scheduling collaborative delay and the upper bound of the collaborative delay. The split quantity adjustment offset amount represents the positive difference between the subtask split quantity adjustment coefficient and the split quantity adjustment setting value. The split quantity adjustment reference amount represents the negative difference between the subtask split quantity adjustment coefficient and the split quantity adjustment setting value.

[0094] In this embodiment, the collaborative task splitting granularity control mechanism achieves coordinated optimization of cross-capacity scheduling efficiency and system resource utilization through precise policy implementation in different scenarios: when the collaborative delay is within the threshold range, no control is implemented to avoid wasting resources on ineffective operations and to maintain the stability of collaborative task processing, ensuring smooth cross-capacity cooperation in multi-scenario delivery and ensuring stable order delivery rhythm; when the collaborative delay exceeds the upper limit, a targeted strategy is formulated based on the actual situation using the precise process of "averaging the delay and computing power correction amount - splitting coefficient determination": if the splitting coefficient exceeds the set value, the granularity is increased to achieve parallel and efficient processing of sub-tasks, shorten the time of collaborative instructions, alleviate scheduling delay, and ensure efficient order delivery; if the splitting coefficient does not reach the set value, the granularity is decreased to reduce the communication overhead and coordination cost of sub-tasks, avoid excessive distribution of computing power, and achieve a dynamic balance between scheduling efficiency and computing power utilization while ensuring the processing quality of core tasks, ensuring stable and timely cross-capacity collaborative scheduling in high-concurrency scenarios, and providing strong support for the efficient operation of the multi-scenario intelligent scheduling system for meal delivery.

[0095] As a further explanation, determining whether to perform granularity control of collaborative task splitting also includes:

[0096] If the cross-capacity scheduling coordination delay is less than the lower bound of the coordination delay, the harmonic average of the coordination delay reference and the computing power adaptation correction is input into the granularity adjustment mapping table to obtain the subtask splitting quantity control factor. It is then determined whether the subtask splitting quantity control factor is greater than the splitting quantity control setting value. This precise determination, combined with the low latency advantage and computing power adaptation, dynamically optimizes the task splitting strategy, avoiding resource redundancy caused by excessive splitting. If so, the rounded result of the interaction between the splitting quantity control compensation and the current coordination task splitting granularity is used as the adjusted coordination task splitting granularity. Appropriately coarsening the granularity reduces the number of subtasks, lowers the frequency of cross-capacity coordination and system communication costs, maintains efficient coordination while avoiding computing power waste, releases redundant resources to other core scheduling tasks, and improves the overall system resource utilization. If efficiency is used, then the rounded result of the interaction between the split quantity control reference and the current collaborative task split granularity is used as the adjusted collaborative task split granularity. More refined task splitting can further explore the potential of cross-capacity collaboration, and use the low latency advantage to achieve more accurate matching and parallel processing of sub-tasks. It is especially suitable for scenarios with complex order details and close cooperation among multiple capacities (such as splitting large orders for delivery). It improves the fineness of collaborative scheduling without increasing latency, and ensures delivery service quality and user experience. The collaborative latency reference represents the negative difference between the cross-capacity scheduling collaborative latency and the lower bound of the collaborative latency. The split quantity control compensation represents the positive difference between the sub-task split quantity control factor and the split quantity control setting value. The split quantity control reference represents the negative difference between the sub-task split quantity control factor and the split quantity control setting value.

[0097] In this embodiment, the collaborative task splitting granularity control mechanism is designed for high-efficiency scenarios where cross-capacity scheduling collaboration latency is less than the lower bound. It achieves dual optimization of resource utilization and scheduling refinement: through a precise process of "harmonizing and averaging collaboration latency and computing power correction amount - determining the control factor," the splitting strategy is dynamically optimized by combining the low latency advantage with computing power adaptation, effectively avoiding resource redundancy caused by excessive splitting; if the control factor exceeds the set value, lowering the granularity can reduce the number of sub-tasks, reduce the frequency of cross-capacity coordination and communication costs, and release redundant computing power to core scheduling tasks while maintaining efficient collaboration, thereby improving the overall resource utilization efficiency of the system; if the control factor does not reach the set value, raising the granularity can further tap the collaboration potential, and achieve accurate matching and parallel processing of sub-tasks with the advantage of low latency, especially suitable for complex order splitting and delivery scenarios. It improves the refinement of collaborative scheduling without increasing latency, effectively ensuring the quality of delivery services and user experience, and providing strong technical support for efficient collaboration of multi-scenario meal delivery.

[0098] 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 disclosed 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 multi-scene dining delivery intelligent scheduling management system, characterized in that, The system comprises a multi-source road condition fusion perception module, a fusion effectiveness intelligent regulation module, a high-concurrency computing power evaluation module and a computing power adaptability intelligent regulation module. The multi-source road condition fusion perception module is configured to obtain multi-source road condition data fusion effectiveness evaluation based on multi-source heterogeneous road condition data, and reflect the quantitative degree of cross-scene road condition information synchronization accuracy and path planning data effectiveness. The fusion effectiveness intelligent regulation module is configured to determine whether to perform fusion effectiveness intelligent regulation according to the multi-source road condition data fusion effectiveness evaluation. If yes, the updated fusion road condition information is generated after regulation and transmitted to the high-concurrency computing power evaluation module. If no, the updated fusion road condition information is directly generated and transmitted to the high-concurrency computing power evaluation module. The fusion effectiveness intelligent regulation comprises time granularity regulation and cross-validation node number regulation. The high-concurrency computing power evaluation module is configured to obtain high-concurrency computing power data in a high-concurrency scene, and obtain high-concurrency scene computing power adaptability evaluation based on the high-concurrency computing power data. The high-concurrency computing power adaptability evaluation is configured to reflect the quantitative degree of delivery timeliness and meal delivery intelligent scheduling adaptability in the high-concurrency scene. The computing power adaptability intelligent regulation module is configured to determine whether to perform computing power adaptability intelligent regulation according to the high-concurrency scene computing power adaptability evaluation. If yes, the meal delivery intelligent scheduling instruction is sent after regulation. If no, the meal delivery intelligent scheduling instruction is directly sent. The computing power adaptability intelligent regulation comprises low-priority instruction cache queue length threshold regulation and collaborative task splitting granularity regulation. The specific steps of determining whether to perform fusion effectiveness intelligent regulation are as follows: If the multi-source road condition data fusion effectiveness evaluation is higher than or equal to the effectiveness set value, fusion effectiveness intelligent regulation is not performed. If the multi-source road condition data fusion effectiveness evaluation is lower than the effectiveness set value, it is determined whether to perform time granularity regulation according to the multi-source road condition data heterogeneous frequency. If yes, it is determined whether to perform cross-validation node number regulation after regulation. If no, it is directly determined whether to perform cross-validation node number regulation. The specific determination steps of determining whether to perform cross-validation node number regulation are as follows: If the data cache preloading time window is greater than the time window reference upper limit, the arithmetic average of the time window correction amount and the effectiveness correction amount is input into the time window mapping table for index query to obtain a verification node number reduction amount. The current double-source cross-validation node number is reduced by the verification node number reduction amount to obtain an adjusted double-source cross-validation node number. The time window correction amount is configured to reflect the positive deviation degree of the data cache preloading time window and the time window reference upper limit. If the data cache preloading time window is within the time window reference interval, cross-validation node number regulation is not performed. The time window reference interval represents a closed interval formed by the time window reference lower limit and the time window reference upper limit. If the data cache preloading time window is less than the lower limit of the time window reference, the time window contrast amount and the validity correction amount are arithmetically averaged and the result is input into the time window mapping table for index query to obtain the verification node quantity increase amount, and the current double-source cross-verification node quantity is coupled with the verification node quantity increase amount to obtain the adjusted double-source cross-verification node quantity. The time window contrast amount is used to reflect the negative deviation degree of the data cache preloading time window from the upper limit of the time window reference.

2. The multi-scene dining distribution intelligent scheduling management system according to claim 1, characterized in that, The multi-source heterogeneous road condition data includes cross-regional road condition information synchronization delay, cross-source data time sequence alignment error, and road network node state perception mismatch frequency; The synchronization delay influence factor and the synchronization delay influence score are interactively processed to obtain a synchronization delay influence value. The synchronization delay influence score represents the result of proportion analysis of the synchronization delay reference value and the cross-regional road condition information synchronization delay. The alignment error influence factor and the alignment error influence score are interactively processed to obtain an alignment error influence value. The alignment error influence score represents the result of proportion analysis of the alignment error reference value and the cross-source data time sequence alignment error. The mismatch frequency influence factor and the mismatch frequency influence score are interactively processed to obtain a mismatch frequency influence value. The mismatch frequency influence score represents the result of proportion analysis of the mismatch frequency reference value and the road network node state perception mismatch frequency. The synchronization delay influence value, the alignment error influence value, and the mismatch frequency influence value are superimposed to obtain a multi-source road condition data fusion effectiveness evaluation.

3. The multi-scene dining distribution intelligent scheduling management system according to claim 2, characterized in that, The specific judgment steps for determining whether to perform time granularity regulation according to the multi-source road condition data heterogeneous frequency are as follows: If the multi-source road condition data heterogeneous frequency is higher than the upper limit of the heterogeneous frequency reference, the heterogeneous frequency correction amount is input into the granularity mapping table for index query to obtain a granularity adjustment factor. It is determined whether the granularity adjustment factor is greater than the granularity adjustment reference value. The heterogeneous frequency correction amount is used to reflect the positive deviation degree of the multi-source road condition data heterogeneous frequency from the upper limit of the heterogeneous frequency reference. If yes, the result of harmonic mean of the granularity adjustment correction amount and the effectiveness correction amount is input into the granularity mapping table for index query to obtain a granularity down-regulation amount. The current dynamic interpolation time granularity is difference processed with the granularity down-regulation amount to obtain an adjusted dynamic interpolation time granularity. The granularity adjustment correction amount is used to reflect the positive deviation degree of the granularity adjustment factor from the granularity adjustment reference value. The effectiveness correction amount is used to reflect the negative deviation degree of the multi-source road condition data fusion effectiveness evaluation from the effectiveness setting value. If no, the result of harmonic mean of the granularity adjustment reference amount and the effectiveness correction amount is input into the granularity mapping table for index query to obtain a granularity up-regulation amount. The current dynamic interpolation time granularity is coupled with the granularity up-regulation amount to obtain an adjusted dynamic interpolation time granularity. The granularity adjustment reference amount is used to reflect the negative deviation degree of the granularity adjustment factor from the granularity adjustment reference value.

4. The multi-scene dining distribution intelligent scheduling management system according to claim 3, characterized in that, The determination of whether to perform time granularity regulation also includes: If the heterogeneous frequency of the multi-source road condition data is within the heterogeneous frequency reference interval, the heterogeneous frequency is input into the granularity mapping table for index query to obtain a granularity control factor, and it is judged whether the granularity control factor is greater than a granularity control reference value, wherein the heterogeneous frequency reference interval represents a closed interval formed by a heterogeneous frequency reference lower limit and a heterogeneous frequency reference upper limit; If yes, the result of harmonic mean of the granularity control reference amount and the effectiveness correction amount is input into the granularity mapping table for index query to obtain a granularity increase amount, and the current dynamic interpolation time granularity is coupled with the granularity increase amount to obtain an adjusted dynamic interpolation time granularity, wherein the granularity control reference amount is used to reflect the positive deviation degree of the granularity control factor and the granularity control reference value; If no, the result of harmonic mean of the granularity control correction amount and the effectiveness correction amount is input into the granularity mapping table for index query to obtain a granularity decrease amount, and the current dynamic interpolation time granularity is coupled with the granularity decrease amount to obtain an adjusted dynamic interpolation time granularity, wherein the granularity control correction amount is used to reflect the negative deviation degree of the granularity control factor and the granularity control reference value; If the heterogeneous frequency of the multi-source road condition data is lower than the heterogeneous frequency reference lower limit, no time granularity control is performed.

5. The multi-scene dining distribution intelligent scheduling management system according to claim 1, characterized in that, The high-concurrency computing power data includes multi-source road condition data fusion effectiveness evaluation, concurrent scheduling instruction processing throughput, and cross-transportation scheduling collaborative delay; The effectiveness evaluation influence value is obtained by interactive processing of the fusion effectiveness influence factor and the effectiveness evaluation influence score, wherein the effectiveness evaluation score represents the result of proportion analysis of the multi-source road condition data fusion effectiveness evaluation and a fusion effectiveness reference value; The throughput influence value is obtained by interactive processing of the throughput influence factor and the throughput influence score, wherein the throughput influence score represents the result of proportion analysis of the concurrent scheduling instruction processing throughput and a throughput reference value; The collaborative delay influence value is obtained by interactive processing of the collaborative delay influence factor and the collaborative delay influence score, wherein the collaborative delay influence score represents the result of proportion analysis of a collaborative delay reference value and the cross-transportation scheduling collaborative delay; The high-concurrency scenario computing power adaptability evaluation is obtained by superimposed processing of the effectiveness evaluation influence value, the throughput influence value, and the collaborative delay influence value. The specific steps of judging whether to perform the computing power adaptability intelligent control are as follows: If the high-concurrency scenario computing power adaptability evaluation is greater than or equal to an adaptability setting value, no computing power adaptability intelligent control is performed; If the high-concurrency scenario computing power adaptability evaluation is less than the adaptability setting value, it is judged whether to perform low-priority instruction cache queue length threshold control according to the number of meal delivery orders in the current preset unit time, if yes, it is judged whether to perform collaborative task splitting granularity control after the control, and if no, it is directly judged whether to perform collaborative task splitting granularity control.

6. The multi-scene dining distribution intelligent scheduling management system according to claim 5, characterized in that, The specific steps of performing the low-priority instruction cache queue length threshold control are as follows: If the meal delivery order quantity in the preset unit time is greater than the upper limit of the order quantity setting, the concurrent pressure value is input into the cache threshold mapping table for index query to obtain a low-priority instruction cache queue length threshold adjustment factor, and whether the low-priority instruction cache queue length threshold adjustment factor is greater than a threshold adjustment reference value is judged, the concurrent pressure value being used to reflect a positive deviation degree of the meal delivery order quantity in the preset unit time from the upper limit of the order quantity setting; If yes, a result of harmonic mean of the threshold adjustment correction amount and the computing power adaptation correction amount is input into the cache threshold mapping table for index query to obtain a threshold increase amount, and the current low-priority instruction cache queue length threshold and the threshold increase amount are superimposed to obtain an adjusted low-priority instruction cache queue length threshold, the threshold adjustment correction amount being used to reflect a positive deviation degree of the low-priority instruction cache queue length threshold adjustment factor from the threshold adjustment reference value, and the computing power adaptation correction amount being used to reflect a negative deviation degree of the high-concurrency scenario computing power adaptability evaluation from the adaptability setting value; If no, a result of harmonic mean of the threshold adjustment reference amount and the computing power adaptation correction amount is input into the cache threshold mapping table for index query to obtain a threshold decrease amount, and the current low-priority instruction cache queue length threshold and the threshold decrease amount are subtracted to obtain an adjusted low-priority instruction cache queue length threshold, the threshold adjustment reference amount being used to reflect a negative deviation degree of the low-priority instruction cache queue length threshold adjustment factor from the threshold adjustment reference value.

7. The multi-scene dining distribution intelligent scheduling management system according to claim 6, characterized in that, The execution of the low-priority instruction cache queue length threshold regulation further includes: If the meal delivery order quantity in the preset unit time is within the order quantity setting interval, the low-priority instruction cache queue length threshold regulation is not executed, the order quantity setting interval representing a closed interval formed by the lower limit of the order quantity setting and the upper limit of the order quantity setting; If the meal delivery order quantity in the preset unit time is less than the lower limit of the order quantity setting, a concurrent relaxation degree is input into the cache threshold mapping table for index query to obtain a low-priority instruction cache queue length threshold regulation factor, and whether the low-priority instruction cache queue length threshold regulation factor is greater than a threshold regulation reference value is judged, the concurrent relaxation degree being used to reflect a negative deviation degree of the meal delivery order quantity in the preset unit time from the lower limit of the order quantity setting; If yes, a result of harmonic mean of the threshold regulation reference amount and the computing power adaptation correction amount is input into the cache threshold mapping table for index query to obtain a threshold decrease amount, and the current low-priority instruction cache queue length threshold and the threshold decrease amount are subtracted to obtain an adjusted low-priority instruction cache queue length threshold, the threshold regulation reference amount being used to reflect a positive deviation degree of the low-priority instruction cache queue length threshold regulation factor from the threshold regulation reference value; If not, the result of the harmonic average of the threshold regulation correction amount and the computing power adaptation correction amount is input into the cache threshold mapping table for index query to obtain a threshold increase amount, and the current low-priority instruction cache queue length threshold is superimposed with the threshold increase amount to obtain an adjusted low-priority instruction cache queue length threshold. The threshold regulation correction amount is used to reflect the negative deviation degree of the low-priority instruction cache queue length threshold regulation factor and the threshold regulation reference value.

8. The multi-scene dining distribution intelligent scheduling management system according to claim 5, characterized in that, The specific process of determining whether to perform collaborative task splitting granularity regulation is as follows: If the cross-transportation force scheduling collaboration delay is within the collaborative delay threshold interval, the collaborative task splitting granularity regulation is not performed. The collaborative delay threshold interval represents a closed interval formed by the collaborative delay lower bound and the collaborative delay upper bound. If the cross-transportation force scheduling scheduling scheduling collaboration delay is greater than the collaborative delay upper bound, the result of the harmonic average of the collaborative delay correction amount and the computing power adaptation correction amount is input into the granularity adjustment mapping table for query to obtain a subtask splitting number adjustment coefficient. It is determined whether the subtask splitting number adjustment coefficient is greater than a splitting number adjustment setting value. If yes, the upper integer result of the interaction between the splitting number adjustment offset amount and the current collaborative task splitting granularity is taken as the adjusted collaborative task splitting granularity. If not, the lower integer result of the interaction between the splitting number adjustment reference amount and the current collaborative task splitting granularity is taken as the adjusted collaborative task splitting granularity. The collaborative delay correction amount represents the positive deviation degree of the cross-transportation force scheduling collaboration delay and the collaborative delay upper bound. The splitting number adjustment offset amount represents the positive deviation degree of the subtask splitting number adjustment coefficient and the splitting number adjustment setting value. The splitting number adjustment reference amount represents the negative deviation degree of the subtask splitting number adjustment coefficient and the splitting number adjustment setting value.

9. The multi-scene dining distribution intelligent scheduling management system according to claim 8, characterized in that, The determination of whether to perform collaborative task splitting granularity regulation further includes: If the cross-transportation force scheduling collaboration delay is less than the collaborative delay lower bound, the result of the harmonic average of the collaborative delay reference amount and the computing power adaptation correction amount is input into the granularity adjustment mapping table for query to obtain a subtask splitting number regulation factor. It is determined whether the subtask splitting number regulation factor is greater than a splitting number regulation setting value. If yes, the lower integer result of the interaction between the splitting number regulation compensation amount and the current collaborative task splitting granularity is taken as the adjusted collaborative task splitting granularity. If not, the upper integer result of the interaction between the splitting number regulation reference amount and the current collaborative task splitting granularity is taken as the adjusted collaborative task splitting granularity. The collaborative delay reference amount represents the negative deviation degree of the cross-transportation force scheduling collaboration delay and the collaborative delay lower bound. The splitting number regulation compensation amount represents the positive deviation degree of the subtask splitting number regulation factor and the splitting number regulation setting value. The splitting number regulation reference amount represents the negative deviation degree of the subtask splitting number regulation factor and the splitting number regulation setting value.

Images