Leg running service task execution method and system combined with semiconductor logic control, cloud platform and chip

Abstract

The invention relates to the technical field of task scheduling and execution, and discloses a semiconductor logic control combined leg running service task execution method and system, a cloud platform and a chip, and the method comprises the steps: receiving a leg running task request submitted by a user, generating a standardized task parameter set, inputting the standardized task parameter set into a pre-trained task execution optimization model, and predicting an optimal task execution scheme; matching candidate nodes through task demand portraits, generating a leg-running node task completion probability model by using the first leg-running node state data flow, and screening out an optimal leg-running node; after the tasks are completed, the completion degree is evaluated according to task execution state data, and fine management of the tasks is achieved; according to the method, the task scheduling efficiency and accuracy are improved, and resource allocation is optimized; the semiconductor logic control is quickly responded through hardware, the real-time performance of the system and the data processing efficiency are improved, manual intervention is reduced, and task execution and settlement are more refined and rationalized.

Description

Technical Field

[0001] The present invention relates to the technical field of task scheduling and execution, and more specifically, to a method, system, cloud platform and chip for executing errand service tasks in combination with semiconductor logic control. Background Art

[0002] With the rapid development of artificial intelligence technology, the task scheduling methods in the field of errand services have increasingly introduced technologies such as machine learning and deep learning. The application of these technologies has greatly improved the execution efficiency and resource utilization of errand tasks. However, in actual applications, there are still problems such as insufficient granularity of task scheduling and lack of real-time performance, which affect the user experience.

[0003] The Chinese patent with the authorization announcement number CN116579339B discloses a task execution method and an optimized task execution method. The method first receives the target description text input by the front end, uses the text processing network to detect whether the key description information conforms to the preset task rules corresponding to the target task, and generates a guiding interaction text if it does not conform to the preset task rules corresponding to the target task, guides the user to feedback the complete description information, and then generates the target task based on the complete interactive description information and executes it. Although this method reduces the threshold and cost of task generation and improves efficiency and explainability, it still lacks optimization and monitoring of the task execution process and cannot achieve refined management of tasks. The Chinese patent application with the publication number CN117055974A discloses an information interaction method and device. After the self-pickup order is generated, the method directly jumps to the errand order page by triggering the call errand control in the self-pickup order details page, and generates the errand order corresponding to the self-pickup order. Although this method combines the self-pickup order with the errand service, reduces the user's operating cost and improves the interaction efficiency, it lacks the screening and management of the errand nodes, and it is difficult to ensure the efficient and high-quality completion of the errand task.

[0004] To sum up, the existing errand running services lack an intelligent screening mechanism for errand running nodes, which affects the quality of task completion; the lack of application of semiconductor technology makes it difficult to achieve real-time response and control of the errand running process. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method, system, cloud platform and chip for executing errand running service tasks in combination with semiconductor logic control, which receives errand running task requests, generates standardized task parameters and predicts the optimal execution plan; intelligently screens the optimal errand running nodes; performs a multi-dimensional evaluation of the task completion degree after the errands are completed, and introduces semiconductor logic control to improve the system real-time performance and processing efficiency, thereby significantly improving the defects of the existing errand running services.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The method for executing a errand service task combined with semiconductor logic control includes:

[0008] Receiving an errand task request submitted by a user, generating a second standardized task parameter set, inputting the second standardized task parameter set into a pre-trained task execution optimization model, and predicting a first optimal task execution plan;

[0009] Constructing a task requirement profile of a first optimal task execution plan, retrieving candidate nodes matching the task requirement profile; receiving a first errand node status data stream of the candidate node; generating a errand node task completion probability model based on the received first errand node status data stream, and screening out the optimal errand node;

[0010] The first optimal task execution plan is pushed to the optimal errand running node. After the errand running task is completed, the task completion time is collected to determine whether the task has timed out. If it has not timed out, the task execution status data of the errand running process is collected. The completion index is obtained based on the task execution status data, and the errand running task is divided into three completion levels: "fully completed", "preliminarily completed" and "not fully completed" according to the completion index.

[0011] Further, generating the second standardized task parameter set includes:

[0012] Extract key parameters from the errand task request to form an original task parameter set; the original task parameter set includes task type, expected completion time, starting data address, and end data address;

[0013] Encode the task type and convert it into the first task type ID;

[0014] Performing timestamp conversion on the expected completion time to generate a first completion timestamp;

[0015] Geocoding the starting point data address and the end point data address to obtain the first standard data coordinates;

[0016] Aggregate the first task type ID, the first completion timestamp, the first standard data coordinates, etc. to form a first standardized task parameter set;

[0017] Retrieving historical tasks that match the first standardized task parameter set of the current task, and generating a second historical task parameter set;

[0018] Furthermore, generating the second standardized task parameter set also includes:

[0019] Calculate the multi-dimensional difference between the first standardized task parameter set and the second historical task parameter set in terms of task type, expected completion time, starting data address and end data address; the multi-dimensional difference includes task type difference, expected completion time difference, starting data address difference and end data address difference;

[0020] Generate a multidimensional deviation vector of the current task relative to the historical task based on the multidimensional difference;

[0021] The first standardized task parameter set is corrected according to the multidimensional deviation vector to obtain a second standardized task parameter set.

[0022] Further, the retrieving the historical tasks matching the first standardized task parameter set of the current task and generating the second historical task parameter set includes:

[0023] Retrieving a task history database based on a first task type ID and a first standard data coordinate in a first standardized task parameter set;

[0024] Get the historical task type ID, calculate the task type similarity between the first task type ID and the historical task type ID. If the two IDs are exactly the same, the task type similarity is 1, otherwise it is 0;

[0025] Obtain the standard data coordinates of the historical tasks, record them as the second standard data coordinates, calculate the Euclidean distance between the first standard data coordinates and the second standard data coordinates, and obtain the data flow similarity;

[0026] Perform weighted summation of task type similarity and data flow similarity to form a comprehensive similarity index;

[0027] Sort the historical tasks according to the comprehensive similarity index, select the top N1 historical tasks with the highest comprehensive similarity index and mark them as historical tasks matching the first standardized task parameter set; where N1≥1.

[0028] Further, generating a running node task completion probability model based on the received first running node status data stream, and screening out the optimal running node includes:

[0029] Constructing an original feature set for running node evaluation according to the first running node state data stream;

[0030] According to the original feature set of the errand node evaluation and the probability model of the errand node task completion, the probability score of each candidate node successfully completing the task is obtained; the candidate nodes are sorted according to the probability score, and the node ranked first is determined as the optimal errand node.

[0031] Further, constructing the original feature set for running errand node evaluation according to the first running errand node state data stream includes:

[0032] Extracting the node position, node idle time and node quality index of the candidate node from the first errand node status data stream;

[0033] Calculate the spatial distance between the node position of the candidate node and the starting point data address in the original task parameter set to obtain the position score of the node relative to the task starting point;

[0034] The idleness score of the node is obtained according to the percentile of the idle time of the node in the idle time distribution of all candidate nodes;

[0035] Calculate the historical score of candidate nodes based on node quality indicators;

[0036] According to the location score, idleness score and historical score of the candidate node, the node evaluation index of the candidate node is obtained;

[0037] The node evaluation indexes of all candidate nodes are aggregated to form the original feature set of errand node evaluation.

[0038] Furthermore, the completion index includes a timeliness index and a completion index;

[0039] The method of dividing the errand task into three completion levels of "completed", "preliminarily completed" and "not fully completed" according to the completion index includes:

[0040] When the completion index falls When the interval is reached, the corresponding level is "all completed";

[0041] When the completion index falls When the interval is reached, the corresponding level is “preliminary completion”;

[0042] When the completion index falls The corresponding level is "Not fully completed" when the interval is reached.

[0043] The system for executing errand service tasks in combination with semiconductor logic control is based on the above-mentioned method for executing errand service tasks in combination with semiconductor logic control, and the system comprises:

[0044] An execution method generation module is used to receive a running errand task request submitted by a user, generate a second standardized task parameter set, input the second standardized task parameter set into a pre-trained task execution optimization model, and predict a first optimal task execution plan;

[0045] Node screening module: used to construct the task requirement profile of the first optimal task execution plan, retrieve candidate nodes matching the task requirement profile; receive the first errand node status data stream of the candidate node; generate the errand node task completion probability model based on the received first errand node status data stream, and screen out the optimal errand node;

[0046] Task completion evaluation module: used to push the first optimal task execution plan to the optimal errand running node. After the errand running task is completed, the task completion time is collected to determine whether the task has timed out. If it has not timed out, the task execution status data of the errand running process is collected; the completion index is obtained based on the task execution status data, and the errand running task is divided into three completion levels: "fully completed", "preliminarily completed" and "not fully completed" according to the completion index;

[0047] The execution method generation module includes a first logic control unit and a second logic control unit;

[0048] The first logic control unit outputs a high level upon receiving a user request to start keyword extraction; after the extraction is completed, it outputs a low level to pass the extraction result to the next unit;

[0049] The second logic control unit internally constructs an inverted index based on the first task type ID and the first standard data coordinates, and realizes fast retrieval through a semiconductor storage array. At the same time, logic gates are used to implement weighted summation. When the comprehensive similarity index exceeds a preset threshold, a high level is output to add the corresponding historical task to the second historical task data set.

[0050] The node screening module includes a third logic control unit;

[0051] When receiving the task requirement portrait, the third logic control unit outputs a high level and starts receiving the first errand node status data stream; and selects the optimal errand node through a comparator according to the probability score of each candidate node.

[0052] The task completion evaluation module includes a fourth logic control unit;

[0053] The fourth logic control unit includes a timer circuit therein, and determines whether the errand running task has timed out by the timer circuit.

[0054] A chip includes the above-mentioned errand service task execution system combined with semiconductor logic control.

[0055] A cloud platform includes the above-mentioned chip, memory, central processing unit and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned errand service task execution method combined with semiconductor logic control is implemented.

[0056] A computer-readable storage medium stores a computer program, which, when executed, implements the above-mentioned method for executing an errand service task combined with semiconductor logic control.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] The present invention receives the errand task request submitted by the user, generates a standardized task parameter set and inputs the pre-trained task execution optimization model, predicts the optimal task execution plan, can comprehensively consider factors such as task type, starting and ending points, and expected completion time, quickly and accurately obtain the best task execution plan, and improves the efficiency and quality of task scheduling; matches the candidate errand nodes according to the task requirement portrait, and uses the errand node status data stream to generate a node task completion probability model, screens out the optimal errand node, and can select the most suitable person to perform the task, avoid resource waste, and improve the success rate of task completion; after the task is completed, the present invention evaluates the completion degree according to the task completion time and status data, subdivides the task into different completion levels, facilitates the subsequent differentiated management for different situations, and realizes the refined management and control of the entire task process; the present invention introduces a dedicated logic control unit in key links such as task matching and node screening, and quickly makes judgments and controls through hardware circuits, which significantly improves the real-time response capability of the system compared to software processing. At the same time, semiconductor storage and logical operations are used to accelerate data processing and reduce computing delays. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 It is a principle flow chart of the method for executing the errand service task combined with semiconductor logic control in the present invention;

[0061] Figure 2 A flow chart of a method for standardizing an original task parameter set in the errand service task execution method combined with semiconductor logic control of the present invention;

[0062] Figure 3 A flow chart of a method for obtaining a second historical task data set matching a first standardized task parameter set in the errand service task execution method combined with semiconductor logic control of the present invention;

[0063] Figure 4A flow chart of a method for constructing an original feature set for evaluating errand nodes in the errand service task execution method combined with semiconductor logic control of the present invention;

[0064] Figure 5 It is a functional module diagram of the errand service task execution system combined with semiconductor logic control in the present invention;

[0065] Figure 6 A schematic diagram of the computer-readable storage medium structure provided for one embodiment of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Example 1

[0068] See also Figure 1 As shown, this embodiment provides a method for executing an errand service task in combination with semiconductor logic control, including:

[0069] Step S1000, receiving the errand task request submitted by the user, and generating a first standardized task parameter set; retrieving historical tasks that match the first standardized task parameter set, generating a second historical task parameter set, generating a second standardized task parameter set based on the first standardized task parameter set and the second historical task parameter set, inputting the second standardized task parameter set into a pre-trained task execution optimization model, and predicting a first optimal task execution plan.

[0070] Furthermore, step S1000 includes:

[0071] Step S1100, receiving the errand task request submitted by the user, extracting key parameters in the errand task request, and forming an original task parameter set; the original task parameter set includes the task type, expected completion time, starting point data address, and end point data address;

[0072] Specifically, users submit errand task requests through mobile application clients or web services. After receiving the request, the service backend first performs semantic analysis on the request content. Semantic analysis uses natural language processing technology to accurately extract key parameters such as task type, expected completion time, starting point and end point data addresses from the unstructured text described by users through methods such as word segmentation, part-of-speech tagging, and named entity recognition. Task types can be preset types such as pickup, delivery, and queuing; the expected completion time is specified by the user or estimated based on the urgency of the task; the starting point data address can be a specific longitude and latitude coordinate, or an address description such as a business name or building number. The extracted key parameters are summarized into the original task parameter set, which records the basic attributes of the task and provides the original input for subsequent task execution optimization. Representing errand needs in a structured way is conducive to the standardized processing and intelligent scheduling of tasks.

[0073] Step S1100 collects errand task requirements in an automated manner, replacing manual communication and information entry, thereby improving the efficiency and accuracy of task acceptance. The application of semantic parsing technology enables intelligent identification of user needs and reduces the risk of misunderstanding due to differences in expression. By refining key parameters, redundant information is eliminated, and a concise and standardized task description form is constructed, which provides a standardized data foundation for the errand service process and is conducive to the optimal design of subsequent links. At the same time, the formation of the original task parameter set also accumulates data samples for task execution analysis, service quality evaluation, etc., and cross-task data mining can further optimize task matching and resource scheduling strategies.

[0074] Step S1200, standardizing the original task parameter set to convert it into a first standardized task parameter set;

[0075] Furthermore, if Figure 2 As shown, step S1200 includes:

[0076] Step S1210, encoding the task type in the original task parameter set and converting it into a first task type ID;

[0077] Step S1220, performing timestamp conversion on the expected completion time in the original task parameter set to generate a first completion timestamp;

[0078] Step S1230, geocoding the start data address and the end data address in the original task parameter set to obtain a first standard data coordinate;

[0079] Step S1240: Aggregate the first task type ID, the first completion timestamp, the first standard data coordinates, etc. to form a first standardized task parameter set.

[0080] Specifically, step S1200 performs standardization on the original task parameter set, the purpose is to eliminate the differences in parameter representation and improve the computability of the parameters. For the task type, it is converted into a digital type identification, i.e., the first task type ID, through a predefined coding rule. For example, the pickup task is coded as 1, the delivery task is coded as 2, and the queuing task is coded as 3. In this way, different task categories can be represented in a unified digital form, which is convenient for the measurement and comparison of task similarities. For the expected completion time, it is converted into a first completion timestamp in the form of a timestamp. The timestamp is the number of seconds or milliseconds relative to a certain reference time point, which can be accurate to a specific moment, providing a quantitative time reference for task progress control and deadline warning. For the starting point and end point data addresses, the geocoding service is used to convert them into the first standard data coordinates under a unified coordinate system. Geocoding can map addresses in various description methods such as address text, points of interest POI, etc. to latitude and longitude coordinates to ensure the consistency of spatial location description. The first standardized task parameter set obtained after standardization provides a unified and standardized data input for subsequent task execution optimization, reducing the processing complexity.

[0081] Step S1200 normalizes the original task parameters of various formats into unified digital indicators through parameter standardization, laying a data foundation for the application of intelligent algorithms. The qualitative description is converted into quantitative characterization, the task type is represented by coding, the time constraint is characterized by timestamp, and the spatial position is described by coordinate system, thus realizing the "digitalization" of the errand demand, so that it can be understood and processed by the computer system. On the one hand, the standardization of parameters is conducive to shortening the time of task similarity calculation and matching retrieval, and improving the optimization efficiency; on the other hand, it is conducive to the statistical analysis of cross-tasks, and by summarizing the regular characteristics of different tasks in various parameters, it can provide ideas for the improvement of optimization algorithms. In addition, the use of standardized parameters also lays a quantitative indicator system for monitoring the task execution process, and intelligent early warning can be realized by setting parameter thresholds. In general, step S1200 is an important foundation for the scientific optimization plan and process automation.

[0082] Step S1300, retrieving historical tasks that match the first standardized task parameter set of the current task, and generating a second historical task parameter set;

[0083] Furthermore, step S1300 includes:

[0084] Step S1310, searching a task history database based on a first task type ID and a first standard data coordinate in a first standardized task parameter set;

[0085] Step S1320, obtaining a second historical task data set matching the first standardized task parameter set;

[0086] Furthermore, if Figure 3 As shown, step S1320 includes:

[0087] Step S1321, obtaining the historical task type ID, calculating the task type similarity between the first task type ID and the historical task type ID, if the two IDs are completely consistent, the task type similarity is 1, otherwise it is 0;

[0088] Step S1322, obtaining the standard data coordinates of the historical tasks, recorded as the second standard data coordinates, calculating the Euclidean distance between the first standard data coordinates and the second standard data coordinates, and obtaining the data flow similarity;

[0089] Step S1323, performing weighted summation on the task type similarity and the data flow similarity to form a comprehensive similarity index;

[0090] Step S1324 , sorting the historical tasks according to the comprehensive similarity index, selecting Top N1 historical tasks with the highest comprehensive similarity index and marking them as historical tasks matching the first standardized task parameter set; wherein N1≥1.

[0091] Specifically, step S1320 adopts a two-stage calculation method to quantitatively evaluate the similarity between the current task and the historical task. In the first stage, targeted similarity measurement functions are designed for the two core features of task type and data flow. For the task type, considering that the requirements of different types of tasks for execution resources vary greatly, a strict consistency matching is adopted, that is, only tasks of exactly the same type are considered similar and assigned a similarity of 1, otherwise 0. Although such similarity calculation is slightly strict, it can ensure that the same type of tasks that have the most direct impact on execution efficiency are selected. For the data flow, considering the continuity of the geographical location, the Euclidean distance is used as the similarity measure, and the smaller the distance, the higher the similarity. The Euclidean distance is the straight-line distance between two points in space, which can characterize the spatial correlation of the starting and ending points of the flow. The closer the distance, the more similar the spatiotemporal constraints on the execution resources. In the second stage, the similarities of the two features are weighted and summed to obtain the comprehensive similarity. The weight coefficient of task type similarity is obtained through statistical analysis of historical data, reflecting the significant impact of task type on execution efficiency; the weight coefficient of data flow similarity is also estimated based on historical data, reflecting the significant impact of data flow on execution efficiency. Based on the single-dimensional similarity, the comprehensive similarity balances and takes into account multi-dimensional features, making the screening of similar tasks more comprehensive and objective. Finally, the comprehensive similarity is used as the sorting basis, and the Top N1 historical tasks are taken as optimization references. The selection of the N1 value takes into account both the computational cost and the optimization benefit, and is generally between 5-10.

[0092] The similarity calculation method of step S1320 can accurately lock the most matching and referenced task samples from massive historical data. The design of task type similarity and data flow similarity fully considers the matching characteristics of the two core features, which is rigorous and meticulous, and operational, and well balances the accuracy and calculation cost of similarity definition. The weighted summation method integrates multi-dimensional similarity, overcomes the one-sidedness that may be brought by a single feature, and makes the screening results more robust and reliable. The setting of Top N1 further controls the number of similar tasks and strikes a balance between the reference value of information and the consumption of computing resources. By quantitatively characterizing similarity, similarity judgment no longer relies on subjective experience, but is based on objective data, which effectively improves the scientific nature of sample screening. At the same time, similarity quantification also provides a unified scale for comparing the similarity between different tasks, so that the correlation between tasks can be characterized by network structure, further expanding the application scenarios of historical experience, and accumulating data materials for algorithm optimization. In addition, since the similarity calculation adopts a highly modular design, each part can be optimized and upgraded independently. For example, the similarity function can introduce more features, the weight coefficient can be estimated using a more sophisticated machine learning model, and the N1 value can be adjusted dynamically, etc., so that the evaluation method of similarity measurement can be continuously iterated and improved, providing more accurate data support for task optimization.

[0093] Step S1330, extracting parameters from the second historical task data to obtain a second historical task parameter set including a second task type ID, a second completion timestamp, and second standard data coordinates;

[0094] Specifically, step S1300 retrieves the second historical task that matches the current task in key features from the massive task history data. First, the first task type ID and the first standard data coordinates in the first standardized task parameter set are used as search conditions to find matching records in the task history database. Among them, the task type and data flow (i.e., data coordinates) are the core features for judging the similarity of tasks. The same task type has similar requirements for execution resources, and similar data flows mean that the time and space constraints are close. Therefore, by determining the search conditions in these two dimensions, historical tasks that match the current task features can be better found. Then, the matching historical task data is taken out to form a second historical task data set. Finally, the data set is subjected to secondary parameter extraction to obtain a second historical task parameter set, including key attributes such as the second task type ID, the second completion timestamp, and the second standard data coordinates. The second historical task parameter set records the historical performance values ​​of a group of similar tasks that match the current task in different dimensions, reflects the execution rules of specific types of tasks, and can provide valuable references for optimization decisions.

[0095] Step S1300 uses the empirical knowledge of historical data to provide intelligent reference for the execution optimization of the current task. Through big data retrieval technology, tasks with matching features are quickly found, and key indicators in their execution records are extracted, so that optimization decisions can be based on previous examples, which reduces decision uncertainty to a certain extent. The completion time distribution of similar tasks can provide a reference for the deadline setting of the current task to avoid overly loose or strict requirements; the data flow characteristics of similar tasks can reveal the distribution law of service resources in different time and space regions, and provide empirical rules for task scheduling; feedback indicators such as execution efficiency and satisfaction of similar tasks can predict the optimization potential of the current task and stimulate the motivation for continuous improvement. At the same time, the acquisition of historical task parameter sets also provides data support for algorithm training. By learning historical experience and summarizing optimization rules, the optimization strategy can continuously iterate itself and continue to maintain its leading position. In general, step S1300 maximizes the value of existing data through similar task mining, which is a beneficial exploration of intelligent optimization.

[0096] Step S1400, comparing and analyzing the deviations between the current task and the historical tasks in terms of task type, expected completion time, starting data address and end data address, and generating a second standardized task parameter set;

[0097] Further, step S1400 includes:

[0098] Step S1410, calculating the multi-dimensional difference between the first standardized task parameter set and the second historical task parameter set in terms of task type, expected completion time, starting data address and end data address; the multi-dimensional difference includes task type difference, expected completion time difference, starting data address difference and end data address difference;

[0099] Step S1420, generating a multidimensional deviation vector of the current task relative to the historical tasks according to the multidimensional difference;

[0100] Step S1430, correcting the first standardized task parameter set according to the multi-dimensional deviation vector to obtain a second standardized task parameter set.

[0101] Specifically, step S1400 uses a quantitative analysis method to evaluate the degree of similarity and difference between the current task and the historical tasks, and accordingly corrects the standardized parameters to make them closer to the actual situation. First, the difference between the first standardized task parameter set and the second historical task parameter set in multiple key dimensions is calculated. The difference reflects the degree of deviation between the two sets of parameters. The greater the difference, the greater the difference between the characteristics of the current task and the historical tasks. Among them, the task type difference describes the difference in the business nature of the two tasks. For example, the difference between picking up and delivering is higher than that of two tasks that are both picking up. The expected completion time difference reflects the difference in the time urgency of the task. The higher the difference, the higher or lower the timeliness requirement of the current task relative to the historical task. The difference between the starting point and end point data addresses reflects the difference in the distribution of tasks in geographical areas. The greater the difference, the lower the overlap of the service radius of the two tasks.

[0102] When calculating the difference, it is necessary to design measurement functions for different parameter data types. For discrete variables such as task types, a simple equality judgment can be used. If they are different, the difference is 1, and if they are the same, it is 0. For continuous variables such as timestamps, they can be measured by ratios or differences of numerical values. For example, the expected completion time difference can be calculated by subtracting the difference between the current task completion timestamp and the average completion timestamp of historical tasks, divided by the standard deviation of the completion timestamps of historical tasks. The larger the difference and the more positive deviations, the more urgent the current task is than similar tasks in the past. Conversely, the urgency is lower. For multidimensional continuous variables such as address coordinates, spatial location differences can be measured by various methods such as Euclidean distance and Manhattan distance. The farther the distance, the higher the difference.

[0103] The result of the difference calculation can be expressed in the form of a multidimensional vector, where each dimension represents a difference index. Next, the difference vectors of the current task relative to all historical tasks are summarized, such as taking the average value, to obtain a comprehensive multidimensional deviation vector. This vector reflects the degree of deviation and direction of deviation of the current task relative to the overall historical level, and is an important basis for correcting the first standardized task parameter set. The specific method of correction is to perform weighted fusion of the first standardized task parameter set and the multidimensional deviation vector to obtain the second standardized task parameter set. The fusion weight coefficient reflects the degree of influence of historical experience on the current decision, which is generally set through experience or learned from data using machine learning methods.

[0104] For example, assuming that the expected completion time of the current pickup task is 2 hours later than the average level of similar historical tasks, the corresponding time dimension deviation is 2, indicating that the urgency of the task is low. Assuming that the weight coefficient of this dimension is 0.5, the completion time in the second standardized task parameter set should be adjusted to be 1 hour later than the completion time of the first standardized task parameter set. Similarly, if the service start and end points of the current pickup task are 10 kilometers away from the average location of historical tasks, the corresponding distance deviation is 10, indicating that the service radius of the task is significantly different from the normal level. Assuming that the weight of this dimension is 0.3, the service coordinates of the second standardized task parameter set should be translated 3 kilometers in the direction of the deviation accordingly. Adjustments in multiple dimensions are performed independently, and ultimately a more realistic revised parameter set is obtained.

[0105] Step S1400 quantitatively analyzes the differences between tasks and modifies the standard parameters accordingly. It can fully adapt to the particularity of the current task without losing historical experience, making the execution plan more personalized and applicable. The design of the difference index comprehensively considers the static attributes and dynamic constraints of the task, comprehensively describes the task characteristics, and greatly enhances the comparability between tasks. The application of multidimensional deviation vectors reflects the balance between experience reference and innovative decision-making, so that the execution optimization strategy better fits the current scenario based on historical reference. At the same time, difference analysis also provides new ideas for task matching, personnel scheduling, etc. For example, when assigning tasks, the undertakers of historical tasks with small deviations from the current tasks can be given priority, thereby improving the success rate of tasks. In dynamic pricing, the pricing strategy can also be adjusted in real time according to the deviation situation to balance the supply and demand relationship. In addition, by tracking and analyzing the time trend of task deviations, the direction of changes in business needs can be predicted, providing decision support for the optimization of platform operation strategies.

[0106] Step S1500: input the second standardized task parameter set into the pre-trained task execution optimization model to predict and obtain the first optimal task execution plan.

[0107] Specifically, the task execution optimization model uses a deep neural network model, which has powerful nonlinear expression capabilities and automatic feature extraction capabilities. It can automatically learn the key factors affecting task execution efficiency from high-dimensional task parameters and establish a complex mapping relationship between input parameters and the optimal execution plan. The network structure of the model includes several stacked fully connected layers, convolutional layers, and recurrent layers, which are used to extract the global features, local features, and time series features of task parameters, respectively. The combination and transformation of features between different layers are completed through activation functions, pooling operations, etc., and finally the probability distribution of the optimal execution plan is output through classifiers such as Softmax.

[0108] The training data set consists of a large number of historical task execution records. Each record includes various parameter information when the task is executed (such as task type, expected completion time, task start and end coordinates, etc.), as well as the actual execution plan and effect evaluation indicators of the task (such as completion time, service quality score, etc.). For each sample, its task parameters are used as model input, and the optimization goal is set to make the model's prediction plan as close as possible to the actual optimal execution plan of the task. By continuously iterating the model parameters and minimizing the prediction error, the model learns to select the most likely successful execution path based on the task attributes. During the model training process, cross-validation and other means can also be used to optimize the generalization performance of the model and improve learning efficiency.

[0109] Step S2000, construct a task requirement portrait of the first optimal task execution plan, retrieve candidate nodes that match the task requirement portrait; receive the first errand node status data stream of the candidate node; generate an errand node task completion probability model based on the received first errand node status data stream, and screen out the optimal errand node.

[0110] Furthermore, step S2000 includes:

[0111] Step S2100, constructing a task requirement profile of a first optimal task execution plan, retrieving a candidate node matching the task requirement profile; receiving a first errand node status data stream of the candidate node;

[0112] Specifically, step S2100 aims to select high-quality candidate nodes with high matching degree with the task and high possibility of completing the task from the massive errand nodes. First, extract the key parameters in the first optimal task execution plan, including task type, expected completion time, start and end points, etc., outline the basic requirements of the task for the execution subject, and form a task demand portrait. Then, take this demand to the errand node resource pool for matching and retrieval, and "recruit troops and horses" through the whole network broadcast to summon qualified errand nodes. In order to avoid resource waste, only nodes with high task attribute matching degree are selected as candidates. The first errand node status data stream includes parameters such as the first real-time data cache status, the first historical execution efficiency, and the first willingness to accept tasks reported by the errand node; among them, the first real-time data cache status reflects the current storage and computing resource occupancy of the node, which is an important basis for judging whether the node is capable of executing a new task; the first historical execution efficiency reflects the performance of the node in completing similar tasks in the past, which can be used as a reference for selecting nodes; the first willingness to accept tasks measures the degree of enthusiasm of the node to accept new tasks from the subjective willingness, and nodes with high willingness are more likely to be conscientious and responsible. These parameter data are collected and summarized in real time in the form of streams to form the first errand node status data stream. Taking into account the hardware status, past performance, subjective attitude and other factors of the node, a multi-dimensional evaluation of its execution potential is conducted to comprehensively examine the availability and reliability of the candidate nodes, providing a solid data foundation for the subsequent optimization of errand nodes.

[0113] Step S2200, generating a running node task completion probability model according to the received first running node status data stream, and screening out the optimal running node;

[0114] Further, step S2200 includes:

[0115] Step S2210, constructing an original feature set for running errand node evaluation according to the first running errand node state data stream;

[0116] Furthermore, if Figure 4 As shown, step S2210 includes:

[0117] Step S2211, extracting the node position, node idle time and node quality index of the candidate node from the first errand node status data stream;

[0118] Step S2212, calculating the spatial distance between the node position of the candidate node and the starting point data address in the original task parameter set, and obtaining the position score of the node relative to the task starting point;

[0119] The method for obtaining the position score of the node relative to the task starting point according to the spatial distance between the node position and the starting point data address is:

[0120]

[0121] in:

[0122] LoP represents the location score, which ranges from 0 to 100. The higher the score, the closer the candidate node is to the starting point of the task and the more suitable it is for the task.

[0123] d represents the Euclidean distance between the candidate node location and the task starting point, in dry meters, which can be obtained by calculating geospatial data.

[0124] k is the adjustment parameter one, k affects the rate of change of the score;

[0125] α is the second adjustment parameter, which affects the smoothness of score changes.

[0126] When the k value is fixed, the larger the α is, the gentler the effect of distance on the score; conversely, the smaller the α is, the more drastic the change of score with distance. The larger the k is, the faster the score decays; the smaller the k is, the slower the score decays.

[0127] The logarithmic structure and power function design of the formula make the position score change nonlinearly, which is closer to people's perception of distance sensitivity. When the distance is close, the score decreases slowly, and as the distance increases, the score decreases at an accelerated rate. The existence of the power exponent α ensures the continuity and smoothness of the score change and avoids step changes. At the same time, the formula output is mapped to a fixed range of 0-100, so that the position scores of different nodes can be compared in a normalized manner. The formula is highly interpretable, efficient in calculation, and flexible in adjustment.

[0128] Step S2213, obtaining an idleness score of the node according to the percentile of the idleness duration of the node in the idleness duration distribution of all candidate nodes;

[0129] Step S2214, calculating the historical score of the candidate node according to the node quality index;

[0130] The method for calculating the historical score of the candidate node according to the node quality index is:

[0131]

[0132] in:

[0133] HiP stands for historical score, which ranges from 0 to 100. A higher score indicates a better historical performance of the candidate node.

[0134] C represents the average time it takes for the candidate node to complete tasks in the past, in minutes.

[0135] C0 is a reference constant of one, which can be set to the average time consumption of the task, reflecting the timeliness of the candidate node’s response to the task. The lower the ratio, the higher the score contribution.

[0136] N represents the total number of tasks that the candidate node has served in the past.

[0137] N0 is the reference constant 2, which can be set to the mean of the total number of platform tasks, reflecting the richness of the candidate node’s service experience. The higher the ratio, the higher the score contribution.

[0138] S represents the average star rating of the candidate node's historical evaluation, ranging from 1 to 5 stars.

[0139] S0 can be set to a full score of 5 stars, reflecting the customer's overall satisfaction with the candidate node's services. The closer the ratio is to 1, the higher the score contribution.

[0140] ω1, ω2, and ω3 are weight coefficients of the three indicators, which control the contribution of each indicator to the total score, and the sum of the weights is 1. It can be adjusted according to actual needs. For example, if timeliness is more important, ω1 can be set larger. Among them, ω1 is the weight of task time, ω2 is the weight of the total task amount, and ω3 is the weight of historical evaluation;

[0141] The formula uses an exponential function. When the indicator performance is better than the reference value, the score increases slowly. When the indicator performance is worse than the reference value, the score decreases rapidly. This design makes the penalty for the decline in service quality greater than the reward for the improvement in performance, guiding the nodes to maintain high-quality services. The introduction of weight coefficients allows the scoring mechanism to flexibly adapt to different task requirements. The formula is concise but contains comprehensive evaluation information, which can accurately reflect the comprehensive historical performance of the node and facilitate comparison and screening.

[0142] Step S2215, obtaining a node evaluation index of the candidate node according to the location score, idleness score and historical score of the candidate node;

[0143] The calculation method of the node evaluation index includes:

[0144]

[0145] in:

[0146] NEI stands for Node Evaluation Index, which ranges from 0 to 100. A higher score indicates that the candidate node is more suitable for the current task overall.

[0147] LoP is the location score of the candidate node.

[0148] HiP is the historical score of the candidate node.

[0149] AvP is the idleness score of the candidate node.

[0150] β1 is the weight index of the position score.

[0151] β2 is the weight index of historical scores.

[0152] β3 is the weight index of the idleness score.

[0153] β1, β2, and β3 are weighted indexes of the three sub-scores, which control their influence on the total index. The sum of the weights is the denominator, so that the NEI finally falls in the range of 0-100. The β value can be set according to the focus of the task. For example, β1 can be set larger in urgent tasks, while β2 can be set larger in tasks with high quality requirements.

[0154] This formula combines three key factors: the location, historical performance, and current status of the node. It adopts the form of weighted geometric mean. The weight β plays a similar role of amplification or reduction of the index. When β is greater than 1, the impact of the indicator is amplified, and when it is less than 1, the impact is reduced. The distribution of β values ​​determines the focus of node selection. The product form of the formula means that when any score is low, the total index will be greatly lowered, reflecting a certain constraint relationship between the various indicators. The selected nodes need to have a relatively balanced performance in all dimensions.

[0155] Compared with the arithmetic mean, the geometric mean can better smooth variables of different dimensions and numerical ranges, and avoid the excessive value of a certain item from having a decisive impact on the overall evaluation. The final square root of the formula makes the numerical distribution of the total index more concentrated, which is convenient for intuitive comparison. In general, the node evaluation index provides a comprehensive and balanced single-indicator evaluation system, which can be used as the main reference for task assignment and dynamic pricing, and improve the accuracy of task matching and the real-time performance of node scheduling.

[0156] Step S2216, summarizing the node evaluation indexes of all candidate nodes to form an original feature set for running node evaluation.

[0157] Specifically, step S2210 extracts three types of key information from the data stream: node location, node idle time, and node quality index. The node location reflects the spatial relationship between the errand node and the place where the task occurs, and directly determines the arrival time cost of the node after receiving the order. The node idle time reflects the current idleness of the errand node. The longer the idle time means that the node is more likely to respond to new tasks first. The node quality index represents the historical service performance of the errand node, mainly including two indicators: average completion time and praise rate. The average completion time measures the service efficiency level of the node. The shorter the time, the stronger the node execution ability. The praise rate measures the service satisfaction of the node. The higher the praise rate means the better the node service quality. These three types of indicators describe the comprehensive service capabilities of the node from the perspectives of spatiotemporal matching, immediacy, and reliability. Since the dimensions and numerical ranges of different indicators vary greatly, in order to facilitate unified processing, the original indicators need to be normalized. Common normalization methods include linear normalization, logarithmic normalization, and probability distribution normalization. Linear normalization makes the indicator dimensionless by linearly mapping the original value to the interval [0,1]. Logarithmic normalization uses the monotonicity of the logarithmic function to map indicators with a large numerical range to a smaller interval. Probability distribution normalization assumes that the indicator follows a certain probability distribution and transforms the indicator value through a probability density function. Different normalization methods are suitable for different scenarios and need to be selected according to the distribution characteristics of the data. After normalization, each indicator is on the same numerical scale, which is convenient for subsequent comprehensive evaluation.

[0158] By calculating the spatial distance between the node position and the task starting point, the position score of the node relative to the task starting point is obtained. The spatial distance can be calculated using common distance measurement functions such as Euclidean distance and Manhattan distance. Euclidean distance is the straight-line distance between two points and is applicable to plane coordinate systems. Manhattan distance is the sum of the absolute wheelbases of two points in the standard coordinate system and is applicable to scenarios such as road networks that are laid out along main roads. The closer the distance, the higher the position score should be. The idleness score of the node is obtained according to the percentile of the idle time of the node in the idle time distribution of all candidate nodes. Percentile is a quantile commonly used in statistics, which indicates the percentage of samples less than or equal to a certain value in the total population. For example, the median is the 50th percentile, which means that 50% of the data items are less than or equal to this value. Quantiles can better reflect the distribution of data. Taking idle time as an example, if the idle time of a node is at the 90th percentile, it means that its idle time is longer than 90% of the nodes, so the idleness is very high. The percentile value of the node can be used directly as the idleness score. The higher the percentile, the higher the score, which means that the node is more idle and should be given priority. The historical score is positively correlated with the trustworthiness of the node in receiving task requests.

[0159] Based on the calculated location score, idleness score, and historical score, geometric weighting is performed to obtain the node evaluation index of the candidate node. The three scores represent the node's capabilities in task matching, order acceptance willingness, and service guarantee. A comprehensive evaluation index is obtained by weighted geometric mean. The weight setting can be adjusted according to actual business needs and platform focus to balance the influence of each evaluation dimension. The obtained node evaluation index can be used as a direct basis for selecting errand nodes. The higher the evaluation index, the stronger the comprehensive service capability of the node and the more suitable it is for the current task. The evaluation indexes of all candidate nodes are aggregated to form a node evaluation index vector, which serves as the original feature set for errand node evaluation. The original feature set contains the evaluation information of all candidate nodes and is an important input for subsequent optimization steps. In the form of a feature set, the service capabilities of candidate nodes can be expressed in a structured form, which is convenient for quantitative analysis and comparison.

[0160] Step S2210 provides a set of scientific and standardized node evaluation methods, which comprehensively considers the various attributes of the errand nodes, so that the errand nodes can be evaluated comprehensively and objectively. By calculating the location score, the node closest to the task location can be given priority, reducing the time for the node to arrive at the task site and improving the efficiency of task completion. By calculating the idleness score, the currently most idle node can be given priority, which increases the node's willingness to accept orders and reduces the probability of task assignment failure. By calculating the historical score, the node with the best performance and the most reliable service in the past can be given priority to improve user satisfaction and task completion quality. Finally, the various evaluation indicators are integrated by weighted summation, making the node evaluation more comprehensive and balanced, and selecting the node with the best comprehensive service capability. At the same time, by constructing a standardized node evaluation original feature set, it is also convenient for the optimization analysis of subsequent steps.

[0161] Step S2220, based on the original feature set of the errand node evaluation and the pre-built errand node task completion probability model, obtain the probability score of each candidate node successfully completing the task; sort the candidate nodes according to the probability score, and determine the node ranked first as the optimal errand node.

[0162] Specifically, the prediction model is constructed by using the ensemble learning paradigm to comprehensively leverage the complementary advantages of multiple base classifiers and improve the accuracy of task completion probability estimation. First, through sampling methods such as bootstrap sampling, some samples are randomly selected from the massive node historical task data to form training sets and test sets. On the one hand, sampling helps to balance the data category distribution and prevent minority class samples from being ignored; on the other hand, dividing the data into two parts can objectively evaluate the generalization ability of the model, which not only provides targeted materials for training, but also reserves a calm and objective scale for testing. Taking the training samples as the learning objects, several base classifier models are trained through classic algorithms such as decision trees and neural networks. During the training process, the original feature set of the errand node evaluation plays the role of input features, providing rich node portrait information, and the final completion status of the corresponding task is used as the target variable, representing the actual control ability of the node for the task. The base classifier summarizes the rules from the big data by learning the correspondence between features and targets, and establishes a causal mapping from the node status to the task results. Through the verification of the test samples, the best is selected from the best, and the best performing base classifier is obtained to prepare for integration.

[0163] The key to ensemble modeling is to design an effective ensemble strategy. According to the number of preferred base classifiers and the degree of homogeneity and heterogeneity, you can choose paradigms such as Stacking, Bagging, and Boosting. For example, when using the Stacking strategy, all base classifiers are first used to predict the samples, and their respective outputs are used as the input of the secondary classifier. After secondary training, the final ensemble probability is obtained. The Bagging strategy uses subsampling to train homogeneous base classifiers with different data sets, and simply averages their prediction results to comprehensively exert individual wisdom. In the Boosting strategy, the base classifiers are generated serially in a sequential manner. The latter continuously strengthens the focus on misjudged samples based on the former, and a group of models with progressive capabilities is obtained after multiple rounds of iterations.

[0164] Regardless of the strategy, it is necessary to use the full amount of historical data as samples, and to set and adjust the parameters of the integrated model to achieve the best overall performance. Model parameters mainly include the combined weights of each base classifier, the number of iterations of the integration process, the amplitude of sample disturbance, etc. Usually, grid search can be used to exhaustively enumerate the parameters and select the parameter combination that minimizes the error of the validation set. It is also possible to use heuristic optimization algorithms such as particle swarms and ant colonies to find the optimal solution for the parameters. After the model training is completed, an integrated estimation model for the completion probability of the errand node task is obtained. This model integrates multiple perspectives, summarizes concise discrimination rules from complex data, and can more accurately predict the control of the node over the task.

[0165] Apply the trained errand node task completion probability model to the actual node selection scenario. When a new errand task needs to be assigned, quickly collect the various evaluation index data of the candidate nodes and convert them into a feature vector form that the model can understand. The feature vector covers all dimensions of the errand node evaluation index system constructed in step S2220, including the node's own attributes, task environment factors, historical performance, etc. Through standardized feature construction, multi-source heterogeneous data from different nodes are uniformly mapped to the same feature space, laying the foundation for fair comparison.

[0166] Next, the feature vector of each candidate node (the original features of the errand node evaluation) will be input into the probability prediction model one by one. Multiple base classifiers work together in an integrated manner inside the model. Each base classifier forms a judgment on whether the node is suitable for performing the task from different perspectives. By means of weighted voting and other means, the opinions of all parties are integrated to generate the probability value of the node successfully completing the task. Due to the adoption of the integration strategy, the errors of individual classifiers are complementary, and the accuracy of group judgment is greatly improved. The probability value is expressed as a real number between 0 and 1, which intuitively quantifies the execution ability of the node and provides a clear standard for the comparison work.

[0167] Based on the probability score of successfully completing the task, the priority of the candidate nodes is determined. The higher the probability score, the greater the probability that the node will successfully complete the task, and the task should be assigned to the node first. By sorting in descending order, the priority of the nodes from high to low is clear at a glance. The node ranked first is the best errand runner recommended by the system. It combines multiple factors such as time, space, and ability, and is most likely to complete the task with high quality and efficiency. Of course, if the probability values ​​of the top several nodes are very close, and it is difficult to choose based on the probability value alone, other decision conditions need to be introduced. Usually, additional attributes such as the distance of the node from the starting point of the task and the praise rate of the node's past services can be considered to perform fine-grained sorting of nodes with similar probability values. The preferred node finally output has a significant advantage in the probability of task execution, and also takes into account the flexibility of actual scheduling, striving to complete the task with the best human resources.

[0168] In actual applications, in addition to the optimal errand node, other high-ranking nodes in the priority list can also be used as high-quality alternative resources. When the best candidate cannot accept the order for some reason, the suboptimal solution can be replaced in turn to ensure the continuity and stability of the errand service. At the same time, nodes with a higher probability of completion can also be included in the standing resource pool, and high-priced orders with higher quality requirements are matched first, thereby improving user experience and expanding profit margins. It can be said that the way step S2240 applies the model to select errand nodes maximizes the potential of human resources, achieves accurate matching of task requirements and node capabilities, and is a key link in intelligent scheduling.

[0169] For example, suppose a running company receives an order to deliver breakfast to Zhang San. It searches for three idle riders A, B, and C nearby, and quickly obtains their real-time status data. Through feature construction, the rider data is converted into a unified indicator vector, which contains multi-dimensional attributes such as the distance between the rider and the merchant, the rider's average order acceptance rate, and the rider's praise last week. After importing the trained task completion probability model, the model outputs the completion probabilities of the three riders A, B, and C as 0.9, 0.8, and 0.7, respectively.

[0170] Since Rider A's probability value is far ahead, Rider A is directly identified as the best candidate and an invitation to accept the order is sent. At the same time, Riders B and C are included in the waiting list as alternatives. Unfortunately, just as Rider A was on his way to the merchant to pick up the meal, his electric bike suddenly ran out of power and broke down, and he could not continue to deliver. After monitoring this situation, the alternative plan was immediately activated, and Rider B, who had the second highest probability value, was notified to take over the task as soon as possible. Since Rider B was close to the merchant and was highly motivated to accept orders, he finally delivered breakfast to Zhang San as scheduled, winning Zhang San's five-star praise.

[0171] Step S3000, push the first optimal task execution plan to the optimal errand running node. After the errand running task is completed, collect the task completion time and determine whether the task has timed out. If it has not timed out, collect the task execution status data of the errand running process; obtain the completion index based on the task execution status data, and divide the errand running task into three completion levels of "fully completed", "preliminarily completed" and "not fully completed" according to the completion index; calculate the reward that each errand running node deserves based on the completion level of the errand running task.

[0172] The completion index includes a timeliness index and a completion index;

[0173] The timeliness index TE is used to measure the timeliness of the completion of the errand task, considering whether the time consumption of the key nodes in the errand process is in line with expectations. The calculation formula of TE is as follows:

[0174]

[0175] in:

[0176] n: the number of key nodes included in the errand task, such as collection, packaging, departure, delivery, and signing, etc.

[0177] i: the i-th key node;

[0178] WTi: The weight of the i-th key node, reflecting the importance of the node in the entire errand running process, and the sum of the weights is 1;

[0179] ETi: expected completion time of the ith key node, predicted based on historical data and road conditions;

[0180] ATi: The actual completion time of the i-th key node, obtained from the task execution status data;

[0181] The value range of TE is (0,1. The purpose of taking the logarithm is to magnify the situation where the actual completion time is lower than the expected time. When the actual completion time of all nodes is equal to the expected time, TE is 1. When the actual time generally exceeds expectations, TE will be significantly lower than 1. The higher the TE value, the more accurate the time control of the errand process.

[0182] The completeness index Cl is used to measure the degree to which goods are delivered completely and information is not missed during the errand process; the calculation formula of Cl is as follows:

[0183] CI=DI·(2-e -SI );

[0184] in:

[0185] DI: Product integrity, reflecting the damage or omission of products during the errand process, with a value range of [0,1]. When the product is intact and the quantity is exactly the same as the order, DI is 1; otherwise, DI will be lower than 1, and the specific value is determined by the proportion of damaged or missing products;

[0186] SI: Information completeness, which reflects the reporting status of key information recorded during the errand running process, and the value range is [0,1]. When the timestamps of the collection, packaging, departure, delivery, and receipt are fully recorded, and the consignee information is correct and the signatures are complete, the SI is 1; otherwise, the SI will be lower than 1, and the specific value is determined according to the weight of the missing information.

[0187] The value range of Cl is (0,2). The purpose of using the exponential form is to amplify the positive contribution of SI to Cl when the value is higher. When the goods and information are complete, Cl will approach 2. If there is a problem in any aspect, Cl will be greatly reduced. The higher the Cl value, the more comprehensive the control of the errand runner over the errand process and the higher the delivery quality.

[0188] The completion index combines the two dimensions of timeliness and completeness, and its calculation formula is:

[0189] The completion index combines the two dimensions of timeliness and completeness, and its calculation formula is:

[0190]

[0191] That is, the completion index is the geometric mean of the timeliness index and the completeness index, and its range is The advantage of this calculation method is that only when the errand running process has good performance in terms of time and quality, can the completion index achieve a high value; if there is an obvious problem in one aspect, even if the other aspect performs well, the completion index will be significantly dragged down. This is conducive to making comprehensive and balanced requirements for errand running services.

[0192] Methods for classifying errand tasks into three completion levels of "fully completed", "preliminary completion" and "not fully completed" based on the completion index include;

[0193] When the completion index falls When the interval is reached, the corresponding level is "All Completed", indicating that all indicators of the errand task have met or exceeded expectations, and the user experience is excellent;

[0194] When the completion index falls In the interval, the corresponding level is “preliminary completion”, indicating that there is still room for improvement in the errand task and the user experience is average;

[0195] When the completion index falls When the interval is reached, the corresponding level is “Not Completed”, indicating that there are obvious problems with the errand task and the user experience is poor;

[0196] The quantitative calculation method of completion index and grade division can accurately reflect the actual completion of errand running tasks. By integrating timeliness and completeness through geometric mean, the completion index can comprehensively consider the key elements of errand running services and balance the importance of time and quality. At the same time, the completion index is directly mapped to three levels, providing a clear and intuitive evaluation scale for errand running tasks. On the one hand, this method can help quantitatively analyze the completion of each task and provide data support for performance appraisal, problem diagnosis, and process optimization; on the other hand, it can also convey a responsible service attitude to users in order to enhance brand reputation and win market trust.

[0197] Specifically, according to the first optimal task execution plan obtained in step S1500, a task notification is pushed to the corresponding errand node through a mobile client or instant messaging tool, including key information such as task type, service start and end locations, expected completion time, and recommended route planning, precautions, and other execution guidelines. After receiving the order, the errand node needs to promptly feedback confirmation information and start executing the task as required.

[0198] During the task execution process, the errand node must regularly upload task execution status data, such as current location, remaining distance, and estimated completion time, through the mobile terminal PDA. On the one hand, these status data can provide users with real-time progress query services and improve service experience; on the other hand, they can also provide a basis for the platform to monitor the progress of task execution and warn of possible timeout risks. The content of task execution status data can be focused on different task types. For example, for pickup tasks, the collection time and cargo photos can be uploaded, and for delivery tasks, delivery certificates and recipient confirmation information can be uploaded.

[0199] After the task is completed, the actual completion time is collected immediately and compared with the expected completion time to determine whether a timeout has occurred. For tasks that have not timed out, the task execution status data of the entire process will be further collected, and a comprehensive completion index will be calculated based on factors such as the actual completion time and user feedback evaluation. This index reflects the performance of the errand node. The higher the index, the better the completion quality. The calculation fully considers multiple dimensions such as timeliness, accuracy and user satisfaction. According to the completion index, errand tasks can be divided into three levels: "completed", "preliminary completion" and "not fully completed". Among them, "completed" means that all aspects of the task are completed as expected, and the time and quality are up to standard; "preliminary completion" means that the core links are basically completed, but there are still deficiencies in details, such as timeouts, poor attitudes, etc.; "not fully completed" indicates that there are problems with missing links or serious non-compliance with the execution standards.

[0200] The results of the completion evaluation will directly affect the incentive returns of the errand nodes. For "fully completed" tasks, the nodes can obtain basic service fees and full rewards; for "preliminary completed" tasks, the nodes can only obtain partial rewards; and for "not fully completed" tasks, the nodes not only cannot obtain rewards, but also need to bear the liability for breach of contract. At the same time, for tasks that have timed out, the corresponding penalty mechanism will be activated. Depending on the severity of the circumstances, penalty measures such as deducting bonuses and restricting orders may be taken, and the service fee will be refunded to the user according to the agreed proportion. The design of the incentive mechanism comprehensively takes into account the results and process of task completion, encourages errand nodes to complete tasks in a timely and high-quality manner, and improves user satisfaction.

[0201] The design of step S3000 fully reflects the concept of process control and equal emphasis on rewards and punishments. First, real-time status data collection allows the backend and users to grasp the progress of tasks in a timely manner, improve process transparency, and provide data support for effective supervision. Users can view task dynamics in real time through the APP, reducing doubts caused by information asymmetry and enhancing trust in errand services. The platform can analyze the task completion trend through status data and intervene in abnormal situations that may affect user experience in advance. At the same time, the task evaluation system based on the completion index has established an objective and fair performance evaluation standard. The index judges the completion of tasks from multiple dimensions such as time, quality, and satisfaction, and can comprehensively measure the level of errand services, facilitating horizontal comparisons across tasks and nodes. The matching incentive mechanism further strengthens the baton role of the index, and guides the errand nodes to consciously improve service quality and strive to complete tasks with positive incentives of rewarding the good and punishing the bad. Transparent evaluation standards and timely reward and punishment feedback can form a virtuous circle and stimulate the service enthusiasm and sense of responsibility of the nodes.

[0202] Example 2

[0203] This embodiment provides, based on the first embodiment, a system for executing errand service tasks combined with semiconductor logic control, such as Figure 5 As shown, including:

[0204] An execution method generation module is used to receive a running errand task request submitted by a user, generate a second standardized task parameter set, input the second standardized task parameter set into a pre-trained task execution optimization model, and predict a first optimal task execution plan;

[0205] Node screening module: used to construct the task requirement profile of the first optimal task execution plan, retrieve candidate nodes matching the task requirement profile; receive the first errand node status data stream of the candidate node; generate the errand node task completion probability model based on the received first errand node status data stream, and screen out the optimal errand node;

[0206] Task completion evaluation module: used to push the first optimal task execution plan to the optimal errand running node. After the errand running task is completed, the task completion time is collected to determine whether the task has timed out. If it has not timed out, the task execution status data of the errand running process is collected; the completion index is obtained based on the task execution status data, and the errand running task is divided into three completion levels: "fully completed", "preliminarily completed" and "not fully completed" according to the completion index.

[0207] The execution method generation module includes a first logic control unit and a second logic control unit;

[0208] The first logic control unit outputs a high level upon receiving a user request to start keyword extraction; after the extraction is completed, it outputs a low level to pass the extraction result to the next unit;

[0209] The second logic control unit internally constructs an inverted index based on the first task type ID and the first standard data coordinates, and realizes fast retrieval through a semiconductor storage array. At the same time, logic gates are used to implement weighted summation. When the comprehensive similarity index exceeds a preset threshold, a high level is output to add the corresponding historical task to the second historical task data set.

[0210] The node screening module includes a third logic control unit;

[0211] When receiving the task requirement portrait, the third logic control unit outputs a high level and starts receiving the first errand node status data stream; and selects the optimal errand node through a comparator according to the probability score of each candidate node.

[0212] The task completion evaluation module includes a fourth logic control unit;

[0213] The fourth logic control unit includes a timer circuit therein, and determines whether the errand running task has timed out by the timer circuit.

[0214] Specifically, the present invention combines a semiconductor logic control unit to realize the automation and intelligence of the errand service task execution process. The first logic control unit is responsible for receiving user requests and starting keyword extraction. When a user submits an errand task request through a mobile terminal or a Web terminal, the first logic control unit detects the request signal, outputs a high-level pulse, and triggers the keyword extraction circuit to start working. Keyword extraction adopts natural language processing (NLP) technology, and realizes parallel computing through semiconductor circuits. It completes word segmentation, part-of-speech tagging, named entity recognition and other processing of the request text within milliseconds, and accurately extracts key parameters such as task type, expected completion time, address coordinates, etc. After the extraction is completed, the first logic control unit outputs a low-level signal to pass the result parameters to the next link.

[0215] Compared with traditional software implementation, this semiconductor keyword extraction has the advantages of fast speed, low power consumption, and high reliability. The parallel computing capability of the dedicated circuit far exceeds that of the general-purpose CPU, and is not limited by the software environment and operating system, and has stronger real-time response capabilities. At the same time, the NLP language model and knowledge base are directly solidified in the chip, without the need for repeated loading, avoiding the memory bandwidth bottleneck. In addition, the power consumption of semiconductor circuits is much lower than that of software and hardware systems, making them very suitable for large-scale, long-term running scenarios. Through the first logic control unit, the system can respond to user requests in a very short time, improve task acceptance efficiency, and simplify subsequent processing through structured representation, laying the foundation for execution optimization.

[0216] The second logic control unit is used to quickly retrieve historical tasks similar to the current task. Its core is a semiconductor storage array built based on task type ID and standard data coordinates. The task type ID uses one-hot encoding, and each bit corresponds to a task type, which can be directly used for content addressing. The standard data coordinates are represented by integers and converted into row and column numbers of the storage array. This design makes full use of the random access characteristics of semiconductor storage and can complete the retrieval of similar tasks in O(1) time. At the same time, the second logic control unit uses logic gate circuits to implement weighted summation and calculate the comprehensive similarity. When the similarity exceeds the preset threshold, the voltage comparator outputs a high level and selects the corresponding historical task into the second historical task data set.

[0217] This semiconductor-based similar task retrieval has the advantages of fast retrieval speed, high storage density, and simple similarity calculation. Traditional software-based retrieval usually uses complex data structures such as inverted indexes, which have high construction and retrieval overhead. Semiconductor arrays can directly map massive amounts of data to physical storage units, realizing storage and computing integration and avoiding data movement. At the same time, weighted summation and threshold comparison are directly implemented using logic gates without complex calculations, and the computing efficiency is much higher than software. In addition, the power consumption of semiconductor circuits is much lower than that of software and hardware systems, making them very suitable for large-scale, long-term running scenarios. Through the second logic control unit, the system can quickly find the historical experience most relevant to the current task, provide valuable reference for execution optimization, and control the scale of calculations through threshold screening to achieve a balance between reference value and computing cost.

[0218] For example, when a user submits a task request to pick up a package from location A and deliver it to location B, the first logic control unit first extracts the task type as "pick up and deliver", the expected completion time as "before 12:00 on November 20, 2022", the coordinates of the pickup location as "(22,114)", and the coordinates of the delivery location as "(23,115)", forming structured task parameters. Then, the second logic control unit uses the key information in the parameters, namely the task type ID (such as "01" for pickup and delivery) and the standardized coordinates (such as "22114" and "23115"), to retrieve historical tasks in the semiconductor storage array in O(1) time. Assume that three similar tasks are retrieved, namely:

[0219] For both pick-up and delivery, the coordinates of the pick-up point are (22,113), and the coordinates of the delivery point are (23,114). The completion time is 10:00 on November 1, 2022.

[0220] For both pick-up and delivery, the coordinates of the pick-up point are (22,114), and the coordinates of the delivery point are (23,116). The completion time is 14:00 on May 2, 2021.

[0221] For both pick-up and delivery, the coordinates of the pick-up point are (21,114), and the coordinates of the delivery point are (23,115). The completion time is 11:00 on March 12, 2022.

[0222] The logic gate circuit calculates the similarity between the three tasks and the current task in the time and space dimensions respectively, and the weighted summation is used to obtain the comprehensive similarity, which is 0.8, 0.7 and 0.85 respectively. Assuming the preset threshold is 0.75, tasks 1 and 3 are selected into the second historical task set, while task 2 is filtered out due to its low time similarity, which has the effect of focusing on relevance.

[0223] The third logic control unit is responsible for triggering the real-time screening process of the errand nodes and selecting the optimal node based on the evaluation results. It is connected to other units or modules through the input / output interface and interacts with the on-chip storage and computing units through control signals and data buses. When the upstream unit generates the task requirement portrait and writes it to the specified storage area, it sends a task ready signal to the third logic control unit. The third logic control unit detects the rising edge of the signal, outputs a high-level pulse, and starts the node screening process. At the same time, it sends a data request to read the static properties and real-time status of the candidate nodes through the on-chip bus, and sends them together with the task portrait to the node evaluation module.

[0224] After the probability scores of all candidate nodes are calculated, the node evaluation module sends an evaluation completion signal to the third logic control unit. The third logic control unit starts to receive a data stream containing node ID and probability score, and selects the node with the highest score through a series of comparators. First, the third logic control unit uses M-way N-bit comparators to compare the N-bit scores of M nodes in parallel. A set of input ends of each comparator is connected to a fixed reference voltage, representing the preset minimum acceptable score, and another set of input ends are connected to the score data lines of M nodes in sequence. If the score of a node is higher than the reference value, the corresponding comparator outputs a high level, indicating that the node is eligible for service. Then, the M-way comparison results are input to the M-select 1 data selector, which is controlled by the priority encoder. The input end of the priority encoder is connected in parallel to the output ends of the M comparators, and after encoding, a log2(M)-bit selection signal is generated to control the M-select 1 data selector to output the node ID with the highest probability score and higher than the minimum acceptable score. Finally, the third logic control unit outputs the selected optimal node ID, which pushes task information to the corresponding node. At the same time, the screening completion signal is fed back to the upstream unit, and it is decided whether to start a new round of screening according to the upstream instruction.

[0225] This semiconductor-based node screening and evaluation has lower response delay. At the same time, the circuit-based implementation also reduces the difficulty of development. Each functional unit can be designed and tested independently, and the combined use greatly shortens the system integration time. The customized data format not only takes into account accuracy and efficiency, but also better fits the characteristics of hardware circuits. Unique hot encoding simplifies the comparison and selection circuits of node IDs, and fixed-point numbers avoid complex floating-point arithmetic units, reducing unit area and power consumption. Based on comparator threshold screening and priority encoding, the optimal node can be found directly without the need to sort all nodes in advance, further reducing the amount of calculation.

[0226] Through the third logic control unit, the errand service platform can complete node screening within milliseconds of receiving a new order, and immediately start task scheduling, making the response time infinitely close to manual dispatching. At the same time, the regularization and parameterization of the screening process allow the scheduling strategy to be flexibly configured and easy to adapt to different business scenarios. Compared with relying on riders to grab orders or unified scheduling, the present invention achieves targeted personalized push through hardware acceleration, which can fully mobilize the service enthusiasm of the errand nodes and better balance supply and demand.

[0227] The fourth logic control unit focuses on the completion evaluation of the errand task. It uses the built-in timer circuit to record the start time, expected completion time, and current time of the task in real time, calculates the remaining time through the semiconductor subtraction circuit, and compares it with the preset threshold voltage to promptly determine the timeout risk and output a warning signal. At the same time, it also receives task status updates from the errand node, determines whether the task is completed through the state machine circuit, and triggers subsequent user confirmation, evaluation collection, and other links.

[0228] This semiconductor task evaluation has the advantages of real-time efficiency, simple customization, and high integration. Traditional software methods make it difficult to strictly unify the clock references of different terminal devices, which affects the timeliness of the evaluation. The on-chip timer can achieve precise timing through a crystal oscillator and is less affected by the software environment. The subtraction circuit is directly implemented with logic gates, and the calculation speed is much faster than the CPU. The state machine circuit can solidify the complex evaluation logic without repeated development. In addition, each functional unit is highly integrated, and triggering and notification are completed through level signals, realizing the automation of the evaluation process.

[0229] Through the fourth logic control unit, the system can control the progress of massive tasks in real time, perceive and respond to risks in a timely manner, and avoid timeouts that affect user experience. The application of the state machine simplifies the evaluation logic, compresses the development and testing workload, and speeds up the system launch. At the same time, the highly integrated design reduces communication between components, reduces data synchronization overhead, and makes the evaluation process smoother.

[0230] In summary, the present invention uses a semiconductor logic control unit to replace traditional software, reshaping the task execution process of the errand service. Through hardware acceleration of key links, the speed and efficiency of task processing are greatly improved. The streamlined instruction system reduces data handling and compresses the reaction chain. The integrated circuit design avoids frequent communication between modules and reduces the complexity of the system. At the same time, the level control method simplifies human-computer interaction, and the application of logic gates makes the calculation process transparent and controllable. The low power consumption characteristics of the dedicated circuit also enable the system to operate stably for a long time. The hardware-level security features eliminate the risk of hacker intrusion from a physical level. The present invention has created a new model of software and hardware collaboration for errand services, which significantly improves platform efficiency and user satisfaction with extremely low latency, extremely high reliability and security, and has broad commercial application prospects.

[0231] Example 3

[0232] This embodiment discloses a chip, including the errand service task execution system combined with semiconductor logic control as described above.

[0233] Example 4

[0234] This embodiment discloses a cloud platform, including the chip, memory, central processing unit and a computer program stored in the memory and executable on the central processing unit as described above. When the central processing unit executes the computer program, the errand service task execution method combined with semiconductor logic control provided above is implemented.

[0235] Since the cloud platform introduced in this embodiment is a platform used to implement the method for executing the errand service task combined with semiconductor logic control in the embodiment of this application, based on the method for executing the errand service task combined with semiconductor logic control introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation method of the cloud platform of this embodiment and its various variations, so how the cloud platform implements the method in the embodiment of this application is not introduced in detail here. As long as the technical personnel of this field implement the cloud platform used by the method for executing the errand service task combined with semiconductor logic control in the embodiment of this application, it belongs to the scope of protection of this application.

[0236] Example 5

[0237] This embodiment discloses a computer-readable storage medium 600, such as Figure 6As shown, computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the method for executing the errand service task combined with semiconductor logic control according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0238] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0239] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD) or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0240] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0241] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0242] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0243] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0244] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0245] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

[0246] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for executing errand service tasks in combination with semiconductor logic control, characterized in that: The method comprises: Receiving an errand task request submitted by a user, generating a second standardized task parameter set, inputting the second standardized task parameter set into a pre-trained task execution optimization model, and predicting a first optimal task execution plan; Constructing a task requirement profile of a first optimal task execution plan, retrieving candidate nodes matching the task requirement profile; receiving a first errand node status data stream of the candidate node; generating a errand node task completion probability model based on the received first errand node status data stream, and screening out the optimal errand node; The first optimal task execution plan is pushed to the optimal errand running node. After the errand running task is completed, the task completion time is collected to determine whether the task has timed out. If it has not timed out, the task execution status data of the errand running process is collected; the completion index is obtained based on the task execution status data, and the errand running task is divided into three completion levels according to the completion index: "fully completed", "preliminarily completed" and "not fully completed".

2. The method for executing errand service tasks in combination with semiconductor logic control according to claim 1, characterized in that: Generating the second standardized task parameter set comprises: Extract key parameters from the errand task request to form an original task parameter set; the original task parameter set includes task type, expected completion time, starting data address, and end data address; Encode the task type and convert it into the first task type ID; Performing timestamp conversion on the expected completion time to generate a first completion timestamp; Geocoding the starting point data address and the end point data address to obtain the first standard data coordinates; Aggregate the first task type ID, the first completion timestamp, the first standard data coordinates, etc. to form a first standardized task parameter set; Retrieve historical tasks that match the first standardized task parameter set of the current task and generate a second historical task parameter set.

3. The method for executing errand service tasks in combination with semiconductor logic control according to claim 2, characterized in that: Generating the second standardized task parameter set also includes: Calculate the multi-dimensional difference between the first standardized task parameter set and the second historical task parameter set in terms of task type, expected completion time, starting data address and end data address; the multi-dimensional difference includes task type difference, expected completion time difference, starting data address difference and end data address difference; Generate a multidimensional deviation vector of the current task relative to the historical task based on the multidimensional difference; The first standardized task parameter set is corrected according to the multidimensional deviation vector to obtain a second standardized task parameter set.

4. The method for executing errand service tasks in combination with semiconductor logic control according to claim 2, characterized in that: The retrieving the historical tasks matching the first standardized task parameter set of the current task and generating the second historical task parameter set comprises: Retrieving a task history database based on a first task type ID and a first standard data coordinate in a first standardized task parameter set; Get the historical task type ID, calculate the task type similarity between the first task type ID and the historical task type ID. If the two IDs are exactly the same, the task type similarity is 1, otherwise it is 0; Obtain the standard data coordinates of the historical tasks, record them as the second standard data coordinates, calculate the Euclidean distance between the first standard data coordinates and the second standard data coordinates, and obtain the data flow similarity; Perform weighted summation of task type similarity and data flow similarity to form a comprehensive similarity index; Sort the historical tasks according to the comprehensive similarity index, select the top N1 historical tasks with the highest comprehensive similarity index and mark them as historical tasks matching the first standardized task parameter set; where N1≥1.

5. The method for executing errand service tasks in combination with semiconductor logic control according to claim 1, characterized in that: The step of generating a task completion probability model of the running node according to the received first running node status data stream and screening out the optimal running node comprises: Constructing an original feature set for running node evaluation according to the first running node state data stream; According to the original feature set of the errand node evaluation and the probability model of the errand node task completion, the probability score of each candidate node successfully completing the task is obtained; the candidate nodes are sorted according to the probability score, and the node ranked first is determined as the optimal errand node.

6. The method for executing errand service tasks in combination with semiconductor logic control according to claim 5, characterized in that: The step of constructing the original feature set for running errand node evaluation according to the first running errand node state data stream comprises: Extracting the node position, node idle time and node quality index of the candidate node from the first errand node status data stream; Calculate the spatial distance between the node position of the candidate node and the starting point data address in the original task parameter set to obtain the position score of the node relative to the task starting point; The idleness score of the node is obtained according to the percentile of the idle time of the node in the idle time distribution of all candidate nodes; Calculate the historical score of candidate nodes based on node quality indicators; According to the location score, idleness score and historical score of the candidate node, the node evaluation index of the candidate node is obtained; The node evaluation indexes of all candidate nodes are aggregated to form the original feature set of errand node evaluation.

7. The method for executing errand service tasks in combination with semiconductor logic control according to claim 1, characterized in that: The completion index includes a timeliness index and a completion index; The method of dividing the errand task into three completion levels of "completed", "preliminarily completed" and "not fully completed" according to the completion index includes: When the completion index falls When the interval is reached, the corresponding level is "all completed"; When the completion index falls When the interval is reached, the corresponding level is "preliminary completion"; When the completion index falls The corresponding level is "Not fully completed" when in interval.

8. A system for executing errand service tasks in combination with semiconductor logic control, which is based on the method for executing errand service tasks in combination with semiconductor logic control as claimed in any one of claims 1 to 7, characterized in that: The system comprises: An execution method generation module is used to receive an errand task request submitted by a user, generate a second standardized task parameter set, input the second standardized task parameter set into a pre-trained task execution optimization model, and predict a first optimal task execution plan; Node screening module: used to construct the task requirement profile of the first optimal task execution plan, retrieve candidate nodes matching the task requirement profile; receive the first errand node status data stream of the candidate node; generate the errand node task completion probability model based on the received first errand node status data stream, and screen out the optimal errand node; Task completion evaluation module: used to push the first optimal task execution plan to the optimal errand running node. After the errand running task is completed, the task completion time is collected to determine whether the task has timed out. If it has not timed out, the task execution status data of the errand running process is collected; the completion index is obtained based on the task execution status data, and the errand running task is divided into three completion levels according to the completion index: "fully completed", "preliminarily completed" and "not fully completed"; The execution method generation module includes a first logic control unit and a second logic control unit; The first logic control unit outputs a high level upon receiving a user request to start keyword extraction; after the extraction is completed, it outputs a low level to pass the extraction result to the next unit; The second logic control unit internally constructs an inverted index based on the first task type ID and the first standard data coordinates, realizes fast retrieval through a semiconductor storage array, and simultaneously uses a logic gate to realize weighted summation. When the comprehensive similarity index exceeds a preset threshold, a high level is output, and the corresponding historical task is added to the second historical task data set; The node screening module includes a third logic control unit; When receiving the task requirement portrait, the third logic control unit outputs a high level to start receiving the first errand node state data stream; and selects the optimal errand node through a comparator according to the probability score of each candidate node; The task completion evaluation module includes a fourth logic control unit; The fourth logic control unit includes a timer circuit therein, and determines whether the errand running task has timed out by the timer circuit.

9. A chip, characterized in that: It includes the errand service task execution system combined with semiconductor logic control as described in claim 8.

10. A cloud platform, characterized in that: It comprises the chip as claimed in claim 9, a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit, wherein when the central processing unit executes the computer program, it implements the method for executing the errand service task combined with semiconductor logic control as claimed in any one of claims 1 to 7.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed, implements the method for executing an errand service task combined with semiconductor logic control according to any one of claims 1 to 7.

Images