Method and device for adjusting execution strategy and nonvolatile storage medium

Abstract

The invention discloses a method and device for adjusting an execution strategy and a nonvolatile storage medium. The method comprises the steps that detection information of an intelligent agent set is acquired, the intelligent agent set comprises multiple intelligent agents used for executing a target task issued by a target object, the detection information comprises performance parameters of each intelligent agent in a preset detection period, and the detection period is included in an execution period of the target task; an evaluation result of the agent set is determined according to the performance parameters, and the evaluation result is used for indicating whether to adjust an execution strategy adopted by the agent set for executing the target task or not; under the condition that the evaluation result indicates to adjust the execution strategy of the agent set, description information of the target task is obtained, the adjustment result of the execution strategy is determined according to the performance parameters and the description information, and the adjustment result comprises the adjustment result of the independent strategy executed by each agent and the adjustment result of the interaction strategy of the multiple agents.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a method and device for adjusting an execution strategy, and a non-volatile storage medium. Background Art

[0002] In complex environments and task scenarios, there are multiple agents. Intelligent orchestration will first analyze the task objectives. For example, in a logistics warehousing scenario, to move goods from area A to area B, the intelligent orchestration system will evaluate the performance characteristics, location status and other factors of each agent (such as a handling robot), and reasonably assign tasks to the agents based on the above factors, so that the agents can collaborate with each other to complete the task objectives. In actual use, since the processing performance of different agents is different, in the process of task orchestration, it is necessary to pre-detect the processing performance status of each agent, and then perform targeted intelligent allocation of data processing tasks based on the processing performance of different agents. However, in related technologies, the detection of the processing performance status of the agent is static, that is, the performance status of the agents participating in the task execution is only detected before the task starts or after the current task ends. Therefore, there is a technical problem that the execution strategy of the agent cannot be flexibly orchestrated according to the dynamic changes in the agent's performance.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] The embodiments of the present application provide a method and device for adjusting the execution strategy, and a non-volatile storage medium, to at least solve the technical problem of being unable to flexibly arrange the execution strategy of the intelligent agent according to the dynamic changes in the performance of the intelligent agent due to the lack of dynamic detection of the intelligent agent in the related technology.

[0005] According to one aspect of an embodiment of the present application, a method for adjusting an execution strategy is provided, comprising: obtaining detection information of an agent set, wherein the agent set includes multiple agents for executing target tasks issued by a target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task; determining an evaluation result of the agent set based on the performance parameters, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set to execute the target task; in a case where the evaluation result indicates adjusting the execution strategy of the agent set, obtaining description information of the target task, and determining the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each agent, and the adjustment result of the interaction strategy of multiple agents.

[0006] Optionally, the evaluation results of the agent set are determined based on the performance parameters, including: dividing the performance parameters into multiple parameter sets according to the agents corresponding to the performance parameters, wherein the performance parameters contained in each parameter set are used to describe the performance of the same agent; for each agent, determining the target parameter set corresponding to the agent, and using a performance prediction model to process and analyze the data in the target parameter set to obtain a performance prediction result, wherein the performance prediction result is used to indicate the performance of the agent in performing the target task in a future time period, the future time period is later than the current moment, and the performance prediction model is obtained by training a neural network model using historical performance parameters of the agent set when performing tasks at historical moments as training data; and determining the evaluation result based on multiple performance prediction results corresponding to multiple agents.

[0007] Optionally, the target parameter set includes different types of performance parameters, wherein the types of performance parameters include: data processing speed, resource utilization, and accuracy; before using the performance prediction model to process and analyze the data in the target parameter set, it includes: for each target parameter set, arranging multiple performance parameters of the same type according to the generation time to obtain a parameter sequence; for each parameter sequence, using a data processing method corresponding to the parameter sequence to process the parameters in the parameter sequence to obtain performance change trend data, wherein the performance change trend data is used to describe the dynamic changes in the target performance when the intelligent agent performs the target task within a preset detection period, and the target performance is the performance of the corresponding type of the parameter sequence.

[0008] Optionally, the parameter sequence is processed using a data processing method corresponding to the parameter sequence to obtain performance change trend data, including: determining a first type of trend evaluation data by the ratio of the sum of the parameter values of all performance parameters contained in the parameter sequence to the target duration corresponding to the parameter sequence, wherein the first type of trend evaluation data is used to evaluate the stability of the intelligent agent in performing the target task; and determining the generation moment of each performance parameter contained in the parameter sequence, and determining the weight value corresponding to each generation moment; determining a second type of trend evaluation data based on all performance parameters and multiple weight values contained in the parameter sequence, wherein the second type of trend evaluation data is used to evaluate the response speed of the intelligent agent in performing the target task; and determining a third type of trend evaluation data based on the parameter values of all performance parameters contained in the parameter sequence and preset coefficients, wherein the third type of trend evaluation data is used to evaluate the delay of the intelligent agent in performing the target task; and determining the first type of trend evaluation data, the second type of trend evaluation data and the third type of trend evaluation data as performance change trend data.

[0009] Optionally, determining the evaluation result based on multiple performance prediction results corresponding to multiple agents includes: when any performance prediction result indicates that the performance of the agent is abnormal, determining the evaluation result as an execution strategy for adjusting the set of agents.

[0010] Optionally, the descriptive information includes: the task requirements of the target task, including: the expected execution speed; determining the adjustment results of the execution strategy based on the performance parameters and the descriptive information, including: determining the expected resource occupancy of each agent when executing the target task and the expected load of each agent when executing the target task based on the expected execution speed; predicting the predicted performance of each agent when executing the target task with the performance parameters, including: the predicted resource occupancy of each agent and the predicted load of each agent; when the predicted resource occupancy is greater than the expected resource occupancy, obtaining the memory cleaning mechanism of each agent, and adjusting the execution order of each subtask to obtain a sequence adjustment result, wherein the subtask is obtained by splitting the target task; determining the strategy with the memory cleaning mechanism and the sequence adjustment result as the first adjustment result corresponding to the resource occupancy abnormality; when the predicted load is greater than the expected load, reallocating the subtasks to each agent to obtain the task reallocation result; determining the strategy with the task reallocation result as the second adjustment result corresponding to the load abnormality.

[0011] Optionally, predicting the predicted performance of each intelligent agent when executing the target task with the performance parameters includes: determining the influencing factors corresponding to each performance parameter, wherein the influencing factors include at least one of the following: task execution time, holidays, and seasons; determining the performance prediction algorithm corresponding to each type of influencing factor, and using the performance prediction algorithm to obtain multiple performance prediction results for the performance parameters corresponding to the influencing factors; and determining the predicted performance based on the multiple performance prediction results.

[0012] According to another aspect of an embodiment of the present application, a device for adjusting an execution strategy is also provided, including: an acquisition module for acquiring detection information of an agent set, wherein the agent set includes multiple agents for executing target tasks issued by a target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task; an evaluation module for determining an evaluation result of the agent set based on the performance parameters, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set to execute the target task; a determination module for acquiring description information of the target task when the evaluation result indicates adjusting the execution strategy of the agent set, and determining the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each agent, and the adjustment result of the interaction strategy of multiple agents.

[0013] According to another aspect of an embodiment of the present application, a non-volatile storage medium is provided, in which a computer program is stored. The device where the non-volatile storage medium is located executes the above-mentioned method of adjusting the execution strategy by running the computer program.

[0014] According to another aspect of an embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the above-mentioned method for adjusting the execution strategy through the computer program.

[0015] According to another aspect of an embodiment of the present application, a computer program product is further provided, including computer instructions, which implement the above-mentioned method for adjusting the execution strategy when the computer instructions are executed by a processor.

[0016] In an embodiment of the present application, the following is adopted: acquiring detection information of an agent set, wherein the agent set includes a plurality of agents for executing target tasks issued by a target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task; determining an evaluation result of the agent set based on the performance parameters, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set for executing the target task; in the case where the evaluation result indicates adjusting the execution strategy of the agent set, acquiring description information of the target task, and determining the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result Including: the adjustment results of the independent strategy executed by each intelligent agent, the adjustment results of the interactive strategy of multiple intelligent agents, by establishing an intelligent agent continuous evaluation mechanism, periodically performing performance evaluation when the intelligent agent performs different tasks, and obtaining the dynamic performance data of the intelligent agent during the task execution process, providing a basis for dynamically adjusting the execution strategy of the intelligent agent, thereby achieving the purpose of dynamically adjusting the execution strategy of the intelligent agent, thereby realizing the technical effect of improving the stability and execution efficiency of the intelligent agent when performing tasks, and then solving the technical problem of not being able to flexibly arrange the execution strategy of the intelligent agent according to the dynamic changes of the performance of the intelligent agent due to the lack of dynamic detection of the intelligent agent in the relevant technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 is a hardware structure block diagram of a computer terminal for implementing a method for adjusting an execution strategy according to an embodiment of the present application;

[0019] Figure 2 is a flowchart of the steps of a method for adjusting an execution strategy according to an embodiment of the present application;

[0020] Figure 3 It is a structural diagram of a device for adjusting an execution strategy according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings 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 in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:

[0024] Agent: A software program or artificial intelligence entity with autonomous decision-making and learning capabilities, such as a robot or automated equipment.

[0025] In related technologies, the processing performance of agents participating in task execution is only detected before the task begins or after the current task ends. This static detection mechanism cannot timely reflect the dynamic performance changes of the agents during task execution, which may cause the execution strategy programmed for the agents to be inconsistent with actual needs. As a result, there are problems such as unreasonable task allocation, low task execution efficiency, and low stability. To solve this problem, the embodiments of this application provide relevant solutions, which are described in detail below.

[0026] According to an embodiment of the present application, an embodiment of a method for adjusting an execution strategy is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1FIG1 shows a hardware structure block diagram of a computer terminal for implementing a method for adjusting an execution strategy. Figure 1 As shown, the computer terminal 10 may include one or more (illustrated as 102a, 102b, ..., 102n in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0028] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry." The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10. As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0029] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for adjusting the execution strategy in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above-mentioned method for adjusting the execution strategy. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0030] The transmission device 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0031] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0032] The embodiment of the present application provides a method for adjusting the execution strategy that can be run in the above-mentioned operating environment. Figure 2 is a flowchart of the steps of the method for adjusting the execution strategy provided in an embodiment of the present application, such as Figure 2 As shown, the method includes the following steps:

[0033] Step S202, obtaining detection information of the agent set, wherein the agent set includes multiple agents for executing target tasks issued by the target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task.

[0034] The embodiment of the present application provides a method for adjusting execution strategies in complex systems with multiple agents working together. The method provided by the embodiment of the present application is not limited to a single scenario and can be widely applied in various scenarios such as intelligent manufacturing, autonomous driving, cloud computing, etc., significantly enhancing the intelligent management and scheduling capabilities of the agent system. The method provided by the embodiment of the present application detects the performance parameters of the agent (such as computing power, data processing speed, accuracy, resource utilization, etc.) in real time when the agent performs a task, thereby realizing dynamic detection of the agent. In step S202, the dynamic change of the agent's performance within the preset detection period can be determined by obtaining the detection information obtained by detecting the agent during the preset detection period. The performance parameters of the agent will only change dynamically when the agent performs a task. Therefore, the detection period for detecting the dynamic performance parameters of the agent (i.e., the preset detection period) must be included in the execution period of the target task, and the duration corresponding to the preset detection period is less than or equal to the duration corresponding to the execution period of the target task. The above-mentioned target task refers to the task that the agent wants to perform, which is issued by the user (i.e., the target object). The user can issue the task through the terminal device or in the interactive interface. The detection information categorizes agents according to the target tasks they perform. Agents performing the same target task form an agent set, and the dynamic performance parameters generated by the agents contained in each agent set are recorded together. For example, in a logistics and distribution scenario, the agent set consists of multiple handling robots, each responsible for sorting and delivering goods issued by the distribution center's target objects. The preset detection cycle is 1 hour, during which the performance parameters of each agent are continuously monitored, including but not limited to computing power (CPU utilization), data processing speed (sorting speed), model accuracy (accuracy in identifying goods), and resource utilization (battery power).

[0035] Step S204 : determining an evaluation result of the agent set according to the performance parameter, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set for executing the target task.

[0036] After obtaining the performance parameters describing the dynamic performance of the agent in step S202, in step S204, the performance change trend of each agent in a future time period that is a certain length of time away from the current moment is predicted based on the performance parameters of each agent. Based on the performance change of each individual agent, an evaluation is made as to whether to adjust the execution strategy adopted by the set of agents to which it belongs to perform the target task, thereby obtaining an evaluation result. For example, in the logistics and delivery scenario mentioned in the above embodiment, the performance parameters of multiple handling robots performing cargo sorting and delivery are used to predict the performance status of these multiple handling robots in the next delivery peak (i.e., the future time period). Based on the performance status of each handling robot in the future time period, a determination is made as to whether to adjust the handling and delivery strategies of these multiple handling robots.

[0037] According to an optional embodiment of the present application, an evaluation result of an agent set is determined based on performance parameters, including: dividing the performance parameters into multiple parameter sets according to the agents corresponding to the performance parameters, wherein the performance parameters contained in each parameter set are used to describe the performance of the same agent; for each agent, determining a target parameter set corresponding to the agent, and using a performance prediction model to process and analyze the data in the target parameter set to obtain a performance prediction result, wherein the performance prediction result is used to indicate the performance of the agent in performing the target task in a future time period, the future time period is later than the current moment, and the performance prediction model is obtained by training a neural network model using historical performance parameters of the agent set when performing tasks at historical moments as training data; and determining an evaluation result based on multiple performance prediction results corresponding to multiple agents.

[0038] As mentioned in the above embodiment, when a target task is collaboratively executed by multiple agents, each target task corresponds to an agent set. Therefore, when determining whether it is necessary to adjust the execution strategy adopted by the agent set to achieve the target task, it is necessary to comprehensively consider each agent in the agent set. Therefore, when performing the evaluation, it is necessary to evaluate the performance of each agent in executing the target task. In order to evaluate the performance of each agent, in this embodiment, all performance parameters obtained by detecting the agent within a preset detection period are first divided into multiple parameter sets according to the agents corresponding to the performance parameters, wherein each parameter set corresponds to an agent, and the performance parameters in a parameter set are used to describe multiple performances of the same agent; the duration of the above-mentioned preset detection period can be set according to actual conditions, for example, 10 minutes, 30 minutes, or 1 hour. Furthermore, in an embodiment of the present application, a time series prediction model (i.e., a performance prediction model) is used to make short-term predictions on the performance indicators and task execution progress of the agent based on the performance parameters, so as to discover possible performance bottlenecks or task delay risks in advance, thereby determining whether it is necessary to adjust the execution strategy adopted by the agent set to achieve the target task. Therefore, in this embodiment, when evaluating the performance of each agent, the (target) parameter set corresponding to the agent is used as the input data of the performance prediction model. The performance prediction model processes the input data, analyzes the performance change trend of the agent, and further determines the performance of the agent in a period of time (i.e., a future period) later than the current moment and earlier than the completion of the target task based on the performance change trend, thereby obtaining the performance prediction result of the agent. Through the above method, the performance prediction model is used to determine the performance prediction result of each agent in the agent set, and finally determines the evaluation result for indicating whether to adjust the execution strategy based on the multiple performance prediction results corresponding to the agent set. The reason why the above performance prediction model can predict the performance change trend of the agent is that it uses the historical performance parameters of a large number of agents as training data to train the neural network model. The so-called historical performance data is the performance parameters of the agent when performing different tasks at multiple historical moments before the current moment.

[0039] In this embodiment, the performance prediction model can be loaded into the memory. For example, the raw data of the performance prediction model can be loaded from the non-volatile memory into the volatile memory so that the processor can run the performance prediction model. The raw data of the performance prediction model refers to unprocessed data, which generally includes parameters and structural data of the performance prediction model. The structural data can be a calculation relationship based on the parameters, such as the forward propagation calculation relationship between intermediate layers and neurons. Specifically, the structural data can include code related to the structure of the performance prediction model, such as code for executing related calculations between intermediate layers and neurons.

[0040] In one embodiment, a memory area for loading the performance prediction model can be divided, including a structure data storage area and a parameter storage area. The structure data storage area is used to store structure-related code, and the parameters referenced by it can point to the addresses of specific parameters in the parameter storage area through pointers. During the training process of the performance prediction model, parameters may need to be frequently updated, and the parameter values in the parameter storage area can be simply updated.

[0041] Optionally, the target parameter set includes different types of performance parameters, wherein the types of performance parameters include: data processing speed, resource utilization, and accuracy; before using the performance prediction model to process and analyze the data in the target parameter set, it includes: for each target parameter set, arranging multiple performance parameters of the same type according to the generation time to obtain a parameter sequence; for each parameter sequence, using a data processing method corresponding to the parameter sequence to process the parameters in the parameter sequence to obtain performance change trend data, wherein the performance change trend data is used to describe the dynamic changes in the target performance when the intelligent agent performs the target task within a preset detection period, and the target performance is the performance of the corresponding type of the parameter sequence.

[0042] As mentioned in the above embodiments, the performance prediction model is actually a time series prediction model, and its input data should be in the form of a time series. Therefore, in this embodiment, before using the performance prediction model to predict the performance change trend of the intelligent agent, it is necessary to first convert the performance parameters of the intelligent agent into the form of a time series. In addition, the method provided in the embodiment of the present application considers multiple dimensions at the same time when evaluating the performance change trend of the intelligent agent. Therefore, when generating a time series based on a parameter set, the method in this embodiment can be used. Specifically, the (target) parameter set corresponding to each intelligent agent also contains performance parameters that describe different types of performance. In order to achieve multi-dimensional evaluation, multiple performance parameters corresponding to the same type of performance are arranged from early to late according to the generation time of the performance parameters to obtain a parameter sequence. This parameter sequence is related to time and is a time series. Next, the parameters in the parameter sequence are processed using the data processing method corresponding to the parameter sequence to obtain performance change trend data. The performance change trend data can reflect the dynamic changes in the performance of the intelligent agent when performing the target task in the preset detection period. The specific type of performance reflected is related to the data processing method used, because different data processing methods have different focuses when processing data, and the performance reflected by the results is also different; for example, the simple moving average method (SMA) can smooth the data sequence and remove short-term fluctuations. It can be used to process performance parameters to evaluate the stability of the intelligent agent when performing tasks; the weighted moving average method (WMA) gives higher weight values to moments closer to the current moment, which means that it pays more attention to the latest changes in the data. It can be used to process performance parameters to evaluate the response speed of the intelligent agent when performing tasks; the exponential moving average method (EMA) emphasizes the importance of the latest data by giving recent data points an exponentially decreasing weight. It can be used to process performance parameters to evaluate the latency of the intelligent agent when performing tasks.

[0043] According to other optional embodiments of the present application, the parameter sequence is processed using a data processing method corresponding to the parameter sequence to obtain performance change trend data, including: determining a first type of trend evaluation data by the ratio of the parameter values of all performance parameters contained in the parameter sequence and the target duration corresponding to the parameter sequence, wherein the first type of trend evaluation data is used to evaluate the stability of the intelligent agent in performing the target task; and determining the generation moment of each performance parameter contained in the parameter sequence, and determining the weight value corresponding to each generation moment; determining a second type of trend evaluation data based on all performance parameters and multiple weight values contained in the parameter sequence, wherein the second type of trend evaluation data is used to evaluate the response speed of the intelligent agent in performing the target task; and determining a third type of trend evaluation data based on the parameter values of all performance parameters contained in the parameter sequence and preset coefficients, wherein the third type of trend evaluation data is used to evaluate the delay of the intelligent agent in performing the target task; and determining the first type of trend evaluation data, the second type of trend evaluation data and the third type of trend evaluation data as performance change trend data.

[0044] As mentioned in the previous embodiment, the embodiment of the present application evaluates the performance of the intelligent agent from multiple dimensions. When evaluating each performance, different data processing methods are used to process the parameters in the parameter sequence. For example, in this embodiment, the performance types include data processing speed, resource utilization, accuracy, stability, task response speed, task processing delay, etc., and the data processing methods used to evaluate the performance of each of the above types of data are different. For example, when evaluating the stability of the intelligent agent in performing tasks, the simple moving average data processing method is used. According to the formula The limited data processing method processes the performance parameters of the intelligent agent collected within a preset detection period, where t represents the current time point, n is the period of the moving average (i.e., the number of historical data points considered), and P i is the performance index value (i.e., the parameter value of the performance parameter) at the i-th time point, MA t It is the evaluation data used to evaluate the stability of the agent in performing the target task (i.e., the first type of trend evaluation data). For example, the above-mentioned simple moving average method can be used to process the CPU occupancy rate of the detection agent when performing the task within the preset detection period to obtain data for evaluating the stability of the agent in performing the task. For example, the CPU occupancy rate collected in the preset detection period constitutes a parameter sequence P, and a part of the parameter sequence P is in the following form: P6{10%, 12%, 15%, 13%, 18%, 20%}. If the moving average period n=5 is selected, the stability of the agent at the 6th time point is predicted, and the data MA6 for evaluating the stability of the agent at the 6th time point is obtained, then according to the formula The stability of the agent at the sixth time point can be evaluated. Following the above principle, the performance parameters generated by the agent in the preset detection cycle can be processed to predict its stability in executing tasks in the future. When evaluating the response speed of the agent in executing tasks, the weighted moving average data processing method is used. According to the formula The defined data processing method processes the performance parameters collected within the preset detection period to obtain data used to describe the response speed of the intelligent agent in performing tasks (i.e., the second type of trend evaluation data), where w i The weight value of the i-th time point is used to represent the weight value of the performance parameter generated at this time point. The sum of all weight values corresponding to all time points included in the preset detection period is 1, and the weight value meets the following conditions: t-n+1 <w t-n+2 <… <w t When evaluating the delay of the agent in executing the task, the moving exponential average data processing method is used. According to the formula EMA t =α×P t +(1-α)×EMA t-1 The limited data processing method processes the performance parameters of the intelligent agent collected within the preset detection period to obtain data used to describe the delay of the intelligent agent in executing the task (i.e., the third type of trend evaluation data), where α is a preset smoothing coefficient with a value range of 0<α<1, and EMA t-1It is the exponential moving average of the previous time point. In this embodiment, the exponential moving average of the earliest detection moment in the preset detection cycle is set as its parameter value, so EMA1=P1. For example, α=0.3, for the CPU occupancy data in the above example, the exponential moving average is calculated. First, EMA1=P1=10%. Next, calculate EMA2, EMA2=0.3×12%+(1-0.3)×10%=10.6%; continue to calculate EMA3, EMA3=0.3×15%+(1-0.3)×10.6%≈11.92%; and so on, the exponential moving average of the entire parameter sequence can be calculated. After evaluating the multi-dimensional performance of an intelligent agent using different data processing methods as described above, it is possible to evaluate from different dimensions whether there will be abnormalities if the intelligent agent continues to use the current execution strategy to perform the task, so as to further determine whether the current execution strategy needs to be adjusted. For example, to analyze the dynamic performance changes of the processing speed (number of image frames processed per second) of an intelligent agent when performing an image recognition task, and use the moving average method to draw a performance curve. First, the agent's processing speed data is collected at regular intervals (e.g., every 10 seconds) under varying task loads (e.g., image dataset sizes) and environmental variations (e.g., varying hardware configurations) to generate a processing speed time series P. This time series is then processed using simple moving average, weighted moving average, and exponential moving average methods. The simple moving average period is determined, the weights for the weighted moving average are set based on the actual situation (e.g., more recent data are given greater weights; a linearly increasing weighting or other appropriate weight distribution method can be used), and the exponential moving average smoothing coefficient α is set to 0.2. Furthermore, after calculating the moving average for each time point (i.e., the generation moment), a performance curve is plotted with time on the horizontal axis and processing speed (raw data and moving average) on the vertical axis. By comparing the raw data curve with the curves generated by different moving average methods, the long-term trend and short-term fluctuations of the agent's processing speed can be more clearly reflected. For example, if the raw data curve fluctuates significantly, while the simple moving average curve is relatively smooth, this indicates that short-term fluctuations in processing speed within that period have been averaged out, reflecting its long-term average processing level. The weighted moving average curve focuses more on the agent's recent performance and can more quickly reflect the changing trend of processing speed. The exponential moving average curve, while taking into account historical data, pays more attention to recent data and can demonstrate the dynamic changes in performance. The performance curves obtained in this way can be used to evaluate the stability and dynamic performance changes of the agent under different conditions. They can also predict the performance trends of the agent in image recognition tasks in the future, so as to better utilize agent resources or adjust task allocation in orchestration strategies.

[0045] According to some optional embodiments of the present application, an evaluation result is determined based on multiple performance prediction results corresponding to multiple agents, including: when any performance prediction result indicates that the performance of the agent is abnormal, determining the evaluation result as an execution strategy for adjusting the set of agents.

[0046] In this embodiment, after performing a multi-dimensional performance evaluation on each agent in the agent set using the method provided in the above embodiment, the performance change trend of each agent when performing tasks in the future can be predicted. Since when a task corresponds to an agent set, the task is implemented by collaboration of multiple agents in the agent set. Therefore, as long as the performance change trend of an agent in the agent set indicates that there is a possibility of performance abnormality when it performs tasks in the future period, it is considered necessary to adjust the current execution strategy of the agent set; wherein, the performance abnormality of the agent includes: the predicted result of resource occupancy rate is greater than the preset resource occupancy rate, the predicted result of task execution delay is greater than the preset delay, the predicted result of task execution response speed is less than the preset response speed, etc.

[0047] Step S206, when the evaluation result indicates to adjust the execution strategy of the set of intelligent agents, obtain the description information of the target task, and determine the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each intelligent agent, and the adjustment result of the interaction strategy of multiple intelligent agents.

[0048] If the evaluation result obtained by executing step S204 indicates that the execution strategy needs to be adjusted, in step S206, the execution strategy of the agent set that implements the target task is adjusted. Specifically, the strategy of the agent set to implement the target task is re-determined based on the performance parameters of the agent and the description information of the target task. The above-determined strategy of the agent set to implement the target task is the adjustment result of the execution strategy. Under normal circumstances, multiple agents in the agent set collaborate with each other to implement the target task. Therefore, when re-determining the strategy of the agent set to implement the target task, not only the strategy (i.e., independent strategy) adopted by each agent to execute its corresponding subtask needs to be adjusted, but also the interaction strategy of multiple agents needs to be adjusted. Therefore, the adjustment result of the execution strategy includes both the adjustment results of the independent strategy and the adjustment results of the interaction strategy. For example, in the logistics and distribution scenario mentioned in the above embodiment, the cargo sorting task and the delivery task performed by the handling robot are both subtasks of the logistics and distribution task, and the logistics and distribution task is the target task. When adjusting the execution strategy, it is necessary to adjust the independent strategy adopted by the handling robot when independently executing the sorting task, as well as the interaction strategy adopted by multiple handling robots when collaboratively executing the delivery task.

[0049] According to some optional embodiments of the present application, the descriptive information includes: the task requirements of the target task, including: the expected execution speed; determining the adjustment results of the execution strategy based on the performance parameters and the descriptive information, including: determining the expected resource occupancy of each intelligent agent when executing the target task and the expected load of each intelligent agent when executing the target task based on the expected execution speed; predicting the predicted performance of each intelligent agent when executing the target task with the performance parameters, including: the predicted resource occupancy of each intelligent agent and the predicted load of each intelligent agent; when the predicted resource occupancy is greater than the expected resource occupancy, obtaining the memory cleaning mechanism of each intelligent agent, and adjusting the execution order of each subtask to obtain a sequence adjustment result, wherein the subtask is obtained by splitting the target task; determining the strategy with the memory cleaning mechanism and the sequence adjustment result as the first adjustment result corresponding to the resource occupancy abnormality; when the predicted load is greater than the expected load, reallocating the subtasks to each intelligent agent to obtain the task reallocation result; determining the strategy with the task reallocation result as the second adjustment result corresponding to the load abnormality.

[0050] If the evaluation result in step S204 indicates that the execution strategy currently adopted by the agent set needs to be adjusted, the adjusted strategy (i.e., the adjustment result) is determined based on the description information of the task currently executed by the agent set (i.e., the target task) and the current performance parameters of each agent in the agent set; in this embodiment, when determining the adjusted execution strategy, the performance requirements of the task being executed (i.e., the target task) are mainly considered. The performance requirements of the task are contained in the description information of the task, including the requirements for data processing speed (i.e., the expected execution speed), the requirements for execution results (i.e., the expected execution results), and other performance requirements. Of course, the description information of the task also includes other information such as the task type and the task complexity index. In this embodiment, natural language processing technology can be used to perform in-depth semantic analysis on the task requirement description (i.e., the description information), and the core goal of the current task can be determined based on the analysis results. At the same time, a classification algorithm combined with a machine learning model is used to judge the task type, and the complexity of the task is quantified by establishing a complexity evaluation model. In this embodiment, when determining the adjustment results of the execution strategy, it is first determined in which performance dimension the agent has an abnormality. Specifically, the expected indicators of the performance of each dimension when the agent performs the task in order to achieve the task requirements in the description information are first determined. For example, through natural language processing, it is determined that the core goal of the task requirements is execution speed. Then, the multi-dimensional expected performance of the agent when performing the task in order to achieve the execution speed is analyzed, such as the expected resource occupancy rate of each agent when performing the target task and the expected load of each agent when performing the target task. Next, based on the current performance parameters of the agent, the multi-dimensional performance change trend of the agent when performing the task in the future period is determined to obtain the predicted resource occupancy rate of each agent and the predicted load of each agent. If the predicted resource occupancy rate of the agent is greater than the expected resource occupancy rate determined according to the task requirements, it means that the performance of the agent in this dimension is abnormal; if the predicted load is greater than the expected load, it means that the performance of the agent in this dimension is abnormal. After determining in which dimension the agent's performance is abnormal, the strategy adjustment result of the corresponding dimension is further determined. For example, if the task load is too high, the task diversion mechanism is activated to reallocate tasks to each agent, and some tasks (i.e., subtasks) are assigned to other idle agents for execution, or to agents whose actual task load is less than the expected task load and whose actual task load has a large difference from the expected task load. If the agent's performance is abnormal only in the dimension of task load, then the task reallocation result obtained is the adjustment result of the execution strategy of the agent set. However, if in addition to the high task load, there is also an abnormal resource occupancy rate, then the task reallocation result obtained above is only the strategy adjustment result corresponding to the load abnormality (i.e., the second adjustment result). In addition, it is also necessary to determine the strategy adjustment result corresponding to the resource occupancy rate, so that the above task reallocation result and the adjustment strategy corresponding to the resource occupancy rate can be used together as the adjustment result of the execution strategy of the agent set.The above-mentioned resource occupancy rate anomaly may be due to excessive CPU occupancy or insufficient hardware resources. If the hardware resources are about to be exhausted, such as insufficient memory, the agent's memory cleanup mechanism will be triggered or more memory resources will be requested from the system. At the same time, the task execution order will be adjusted to give priority to tasks with less memory requirements, and memory-intensive tasks will be temporarily suspended or transferred to other agents with sufficient memory resources. At this time, the order adjustment results obtained by adjusting the task execution order and the agent's memory cleanup mechanism are both adjustment results corresponding to the resource occupancy rate anomaly (i.e., the first adjustment result). For example, the preset CPU utilization threshold is 0.9. In a set of agents performing image recognition tasks, for each agent, based on performance parameters, its CPU utilization when processing a large amount of image data will increase from the current 0.7 to 0.95 within the next 5 time steps, and the image recognition accuracy will drop from 90% to 70%. Then the execution strategy currently adopted by the agent set needs to be adjusted. In response to the above situation, according to the self-healing strategy, the pre-processing tasks of some image data are assigned to the agent, and the agent focuses on the core image recognition algorithm processing. After adjustment, The CPU utilization of the intelligent agent is stable at around 0.8, and the image recognition accuracy is maintained at above 85%, thus avoiding intelligent agent failures caused by task abnormalities and ensuring the smooth progress of image recognition tasks; in addition, in this embodiment, if the performance of the intelligent agent is judged from the dimension of data processing speed, in addition to setting a processing speed threshold and judging that it is a performance abnormality when it is less than the processing speed threshold, a speed drop threshold (such as 50%) can also be set. When the data processing speed of the intelligent agent drops by more than the drop threshold within a preset time period, the data processing speed performance of the intelligent agent is judged to be abnormal.

[0051] Optionally, predicting the predicted performance of each intelligent agent when executing the target task with the performance parameters includes: determining the influencing factors corresponding to each performance parameter, wherein the influencing factors include at least one of the following: task execution time, holidays, and seasons; determining the performance prediction algorithm corresponding to each type of influencing factor, and using the performance prediction algorithm to obtain multiple performance prediction results for the performance parameters corresponding to the influencing factors; and determining the predicted performance based on the multiple performance prediction results.

[0052] In the previous embodiment, it is mentioned that the dynamic performance changes of the intelligent agent are predicted based on the performance parameters generated by each intelligent agent within a preset detection period. In this embodiment, a dynamic adjustment strategy based on the performance prediction model in the above embodiment is established, that is, a time series prediction model (i.e., a performance prediction model) is used to make short-term predictions on the performance indicators and task execution progress of the intelligent agent, so as to discover possible performance bottlenecks or task delay risks in advance, and thus adjust the scheduling plan in advance. The process of using the time series prediction model here is as follows: Assuming that the performance indicators and task execution progress of the intelligent agent are affected by "trend items", "season items" and "holiday items", the time series composed of performance parameters is decomposed into trend items g(t), season items s(t) and holiday items h(t) and error items ∈ t Therefore, when predicting the dynamic performance changes of the intelligent agent, the influencing factors of each performance parameter are first determined. The influencing factors include: task execution time, holidays, seasons, etc. The type of performance parameter is determined according to the influencing factors corresponding to each performance parameter. Among them, the performance parameter whose influencing factor is task execution time is classified as trend term g(t), the performance parameter whose influencing factor is season is classified as seasonal term s(t), the performance parameter whose influencing factor is holiday is classified as holiday term h(t), and the performance parameter whose influencing factor cannot be determined is classified as error term ∈ t . Performance parameters belonging to trend items include data processing speed, parameters belonging to seasonal items include resource utilization (for example, in the field of education or scientific research, the demand for computing resources at the beginning and end of the semester varies seasonally); performance parameters belonging to holiday items include memory occupancy. In this embodiment, a corresponding performance prediction algorithm is set for each influencing factor, and the performance prediction algorithm corresponding to the influencing factor is used to process the performance parameters corresponding to the influencing factor to obtain a performance prediction result, wherein the performance algorithm used for the trend item g(t) is the algorithm corresponding to the logistic growth model or the algorithm corresponding to the piecewise linear model. For example, the performance prediction algorithm corresponding to the piecewise linear model is the formula The algorithm defined by the algorithm, where g(t) is the performance prediction result of the trend item, which represents the performance index affected by the task execution time at time point t; T1 and T2 represent a turning point in the time series, which is the time point when the performance changes; m and k represent the performance index of the agent at T1 and the performance index of the agent at T2, respectively. t is an indicator function used to distinguish whether the time point t is between T1 and T2. If t is between T1 and T2, then I t =1; if t is not between T1 and T2, then I t = 0; Performance parameters affected by task execution time. The above method is used to predict the trend value of the agent at time point t to predict changes in performance indicators and task progress. The performance prediction algorithm corresponding to the seasonal term s(t) is expressed by the following formula: Where P is the seasonal cycle, a n and b n are the coefficients estimated by minimizing the loss function, where cos represents the cosine function and sin represents the sine function. By using the above formula, the performance parameters affected by seasons are decomposed into Fourier series, which can predict the periodic change pattern of the performance parameters contained in the seasonal terms. The terms corresponding to each Fourier series (n) obtained by decomposition describe fluctuations at a specific frequency. Low-frequency terms (n with small values) represent long-term trends in the data, while high-frequency terms (n with large values) represent fluctuations within a shorter period. Data belonging to holiday items can be filtered by the time frequency at which the data was collected. For example, to perform performance prediction based on a time series consisting of the memory usage of an agent, first determine the time frequency of the data, and filter the data belonging to the holiday items based on the time frequency. For trend items, determine whether to use a logistic growth model or a piecewise linear model based on the overall trend of the data. For example, if the memory usage shows a trend of first slowly increasing and then stabilizing over time, a piecewise linear model may be selected, and k, m, T1, and T2 may be determined based on the data. For the seasonal term, if it is found that the memory usage has a weekly cyclical change (for example, there is a difference between weekends and weekdays), then P = 7, and a is determined by fitting the data to the Fourier series. n and b n Finally, the performance prediction results corresponding to the trend item, the performance prediction results corresponding to the season item, and the performance prediction results corresponding to the holiday item are added together to obtain the predicted value of the performance prediction result representing the intelligent agent. For example, to predict the memory usage rate for the next day, the predicted memory usage rate value is calculated based on factors such as the position of the day in the trend item, the cycle of the seasonal item, and whether it is a holiday.

[0053] It's also important to note that after determining the results of the execution strategy adjustments, each agent is activated according to the adjustments, and communication links and data transmission channels are established between the agents. Simultaneously, the agent's operating status is monitored in real time during task execution. When launching the agents, containerization technology is used to package each agent and its dependent environments into independent containers, allowing for rapid deployment and startup. Furthermore, by embedding performance monitoring probes within the agents and setting up traffic monitoring points in the data transmission channels, performance metrics and task execution progress data are collected in real time.

[0054] Through the above steps, we can monitor the performance parameters of the agent ensemble in real time and, based on the requirements of the target task, dynamically adjust the execution strategy used by the agent ensemble to achieve the target task, ensuring efficient and stable task completion. Using performance prediction models to predict the agent's performance in future time periods can prevent potential performance bottlenecks or anomalies and improve the stability of the agent's task execution. Furthermore, by adjusting the execution strategy, we can effectively optimize resource utilization and improve task execution efficiency.

[0055] Figure 3 is a structural diagram of a device for adjusting an execution strategy according to an embodiment of the present application, such as Figure 3 As shown, the device for adjusting the execution strategy includes: an acquisition module 30, which is used to obtain detection information of an agent set, wherein the agent set includes multiple agents for executing target tasks issued by a target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task; an evaluation module 32, which is used to determine the evaluation result of the agent set based on the performance parameters, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set to execute the target task; a determination module 34, which is used to obtain description information of the target task when the evaluation result indicates to adjust the execution strategy of the agent set, and determine the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each agent, and the adjustment result of the interaction strategy of multiple agents.

[0056] It should be noted that Figure 3 The preferred implementation of the embodiment shown can be found in Figure 2 The relevant description of the illustrated embodiment will not be repeated here.

[0057] An embodiment of the present application further provides a non-volatile storage medium, in which a computer program is stored. The device where the non-volatile storage medium is located executes the above method for adjusting the execution strategy by running the computer program.

[0058] The above-mentioned non-volatile storage medium is used to store programs that perform the following functions: obtaining detection information of an agent set, wherein the agent set includes multiple agents for executing target tasks issued by a target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task; determining an evaluation result of the agent set based on the performance parameters, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set to execute the target task; when the evaluation result indicates to adjust the execution strategy of the agent set, obtaining description information of the target task, and determining the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each agent, and the adjustment result of the interaction strategy of multiple agents.

[0059] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the above method for adjusting the execution strategy through the computer program.

[0060] The processor in the above-mentioned electronic device is used to run a program that performs the following functions: obtaining detection information of an agent set, wherein the agent set includes multiple agents for executing target tasks issued by a target object, and the detection information includes: performance parameters of each agent within a preset detection period, and the detection period is included in the execution period of the target task; determining an evaluation result of the agent set based on the performance parameters, wherein the evaluation result is used to indicate whether to adjust the execution strategy adopted by the agent set to execute the target task; when the evaluation result indicates to adjust the execution strategy of the agent set, obtaining description information of the target task, and determining the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each agent, and the adjustment result of the interaction strategy of multiple agents.

[0061] According to another aspect of an embodiment of the present application, a computer program product is further provided, including computer instructions, which implement the above method for adjusting the execution strategy when the computer instructions are executed by a processor.

[0062] It should be noted that the various modules in the above-mentioned device for adjusting the execution strategy can be program modules (for example, a set of program instructions that implement a certain specific function) or hardware modules. For the latter, it can be expressed in the following forms, but is not limited to this: the expression form of each of the above-mentioned modules is a processor, or the functions of each of the above-mentioned modules are implemented by a processor.

[0063] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0064] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0065] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, 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 through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0066] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0067] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0068] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the relevant technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0069] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for adjusting an execution strategy, characterized in that: include: Acquiring detection information of an agent set, wherein the agent set includes a plurality of agents for executing a target task issued by a target object, and the detection information includes: performance parameters of each of the agents within a preset detection period, wherein the detection period is included in the execution period of the target task; Determining an evaluation result of the agent set according to the performance parameter, wherein the evaluation result is used to indicate whether to adjust an execution strategy adopted by the agent set to execute the target task; When the evaluation result indicates that the execution strategy of the set of intelligent agents should be adjusted, the description information of the target task is obtained, and the adjustment result of the execution strategy is determined based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each of the intelligent agents, and the adjustment result of the interaction strategy of multiple intelligent agents.

2. The method according to claim 1, characterized in that Determining an evaluation result of the agent set according to the performance parameter includes: Dividing the performance parameters into multiple parameter sets according to the agents corresponding to the performance parameters, wherein the performance parameters contained in each parameter set are used to describe the performance of the same agent; For each of the agents, determining a target parameter set corresponding to the agent, and using a performance prediction model to process and analyze data in the target parameter set to obtain a performance prediction result, wherein the performance prediction result is used to indicate the performance of the agent in performing the target task in a future time period, where the future time period is later than the current time, and the performance prediction model is obtained by training a neural network model using historical performance parameters of the agent set when performing tasks at historical time periods as training data; The evaluation result is determined according to a plurality of performance prediction results corresponding to a plurality of the intelligent agents.

3. The method according to claim 2, characterized in that The target parameter set includes different types of performance parameters, wherein the types of performance parameters include: data processing speed, resource occupancy rate, and accuracy rate; Before using the performance prediction model to process and analyze the data in the target parameter set, the following steps are included: For each target parameter set, arranging multiple performance parameters of the same type according to generation time to obtain a parameter sequence; For each of the parameter sequences, the parameters in the parameter sequence are processed using a data processing method corresponding to the parameter sequence to obtain performance change trend data, wherein the performance change trend data is used to describe the dynamic changes in the target performance when the intelligent agent performs the target task within the preset detection period, and the target performance is the performance of the corresponding type of the parameter sequence.

4. The method according to claim 3, characterized in that The parameter sequence is processed using a data processing method corresponding to the parameter sequence to obtain performance change trend data, including: determining first-type trend evaluation data by calculating the ratio of the sum of the parameter values of all the performance parameters included in the parameter sequence to the target duration corresponding to the parameter sequence, wherein the first-type trend evaluation data is used to evaluate the stability of the intelligent agent in performing the target task; and determining a generation time of each of the performance parameters included in the parameter sequence, and determining a weight value corresponding to each of the generation times; determining second-type trend evaluation data based on all the performance parameters included in the parameter sequence and the plurality of weight values, wherein the second-type trend evaluation data is used to evaluate a response speed of the intelligent agent in performing the target task; and Determining third-category trend evaluation data based on parameter values of all the performance parameters included in the parameter sequence and preset coefficients, wherein the third-category trend evaluation data is used to evaluate the delay of the intelligent agent in performing the target task; The first type of trend evaluation data, the second type of trend evaluation data, and the third type of trend evaluation data are determined as the performance change trend data.

5. The method according to claim 2, characterized in that The evaluation result is determined based on multiple performance prediction results corresponding to multiple agents, including: when any one of the performance prediction results indicates that the performance of the agent is abnormal, determining the evaluation result as an execution strategy for adjusting the set of agents.

6. The method according to claim 1, characterized in that The description information includes: the task requirements of the target task, including: expected execution speed; Determining an adjustment result of the execution strategy according to the performance parameter and the description information includes: Determine, based on the expected execution speed, the expected resource occupancy rate of each agent when executing the target task and the expected load of each agent when executing the target task; Predicting the predicted performance of each of the intelligent agents when performing the target task with the performance parameters includes: The predicted resource occupancy rate of each of the intelligent agents and the predicted load of each of the intelligent agents; When the predicted resource occupancy rate is greater than the expected resource occupancy rate, obtaining a memory cleanup mechanism for each agent and adjusting the execution order of each subtask to obtain a sequence adjustment result, wherein the subtask is obtained by splitting the target task; and determining a policy that records the memory cleanup mechanism and the sequence adjustment result as a first adjustment result corresponding to the abnormal resource occupancy rate; When the predicted load is greater than the expected load, the subtasks are reallocated to each of the agents to obtain a task reallocation result; and the strategy recording the task reallocation result is determined as the second adjustment result corresponding to the load anomaly.

7. The method according to claim 6, characterized in that Predicting the predicted performance of each of the intelligent agents when performing the target task with the performance parameters includes: Determining an influencing factor corresponding to each of the performance parameters, wherein the influencing factor includes at least one of the following: task execution time, holidays, and seasons; Determining a performance prediction algorithm corresponding to each type of the influencing factors, and using the performance prediction algorithm to calculate the performance parameters corresponding to the influencing factors to obtain multiple performance prediction results; The predicted performance is determined according to a plurality of the performance prediction results.

8. A device for adjusting an execution strategy, characterized in that: include: an acquisition module configured to acquire detection information of an agent set, wherein the agent set includes a plurality of agents configured to execute a target task issued by a target object, and the detection information includes performance parameters of each of the agents within a preset detection period, wherein the detection period is included in an execution period of the target task; An evaluation module, configured to determine an evaluation result of the agent set based on the performance parameter, wherein the evaluation result is used to indicate whether to adjust an execution strategy adopted by the agent set to execute the target task; A determination module is used to obtain the description information of the target task when the evaluation result indicates that the execution strategy of the set of intelligent agents should be adjusted, and determine the adjustment result of the execution strategy based on the performance parameters and the description information, wherein the adjustment result includes: the adjustment result of the independent strategy executed by each of the intelligent agents, and the adjustment result of the interaction strategy of multiple intelligent agents.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a computer program, wherein the method for adjusting the execution strategy according to any one of claims 1 to 7 is executed by running the computer program on the device where the non-volatile storage medium is located.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method for adjusting the execution strategy according to any one of claims 1 to 7 through the computer program.

11. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the method for adjusting the execution strategy according to any one of claims 1 to 7 is implemented.

Images