Dynamic priority-based integrated soc industrial personal computer real-time scheduling method, system and equipment and medium

Abstract

The invention relates to an integrated soc industrial personal computer real-time scheduling method, system and device based on dynamic priority and a medium. The method comprises the following steps: collecting system state data of an integrated soc industrial personal computer, determining a potential available resource set by adopting a state quantification method based on the resource state data, and evaluating task priorities in combination with a dynamic priority evaluation rule to generate a task priority sequence; generating a preliminary resource allocation-scheduling path scheme according to a preset state space and an action space; deducing a resource allocation and scheduling path adjustment process by simulating a preset emergency scene to obtain an emergency scene simulation result; optimizing the preliminary scheme by adopting a priority-based compensation strategy to form an optimized resource allocation-scheduling path scheme; and generating a scheduling instruction to drive the integrated soc industrial personal computer to execute the task. By adopting the method, the real-time performance and the self-adaptive capability of resource scheduling can be improved, fluctuations caused by emergency scenes can be effectively dealt with, key tasks are ensured to be executed preferentially, and the overall efficiency and reliability of a system are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of industrial automation, and particularly relates to a dynamic-priority-based integrated SOC industrial computer real-time scheduling method, system, device and medium. BACKGROUND

[0002] With the rapid development of the field of industrial automation and intelligent manufacturing, the complexity of industrial control systems is continuously improving, and technical features such as multi-task concurrent execution, close interconnection of devices, and dynamic changes in business scenarios are emerging. The industrial computer of the integrated system on chip (SOC) (referred to as integrated SOC industrial computer) gradually becomes a commonly used core control unit in the industrial control field due to its high integration and low power consumption. The real-time performance and reliability of the task scheduling of the integrated SOC industrial computer directly affect the stability and efficiency of the entire industrial production process. In the traditional technology, a fixed priority scheduling algorithm or a static resource allocation strategy is usually used. The priority level of each task is pre-set, and the resources such as calculation, storage, and communication are allocated in a fixed order. Once the resource allocation amount and the scheduling path are determined, they will not be dynamically adjusted, so as to ensure the basic task execution order.

[0003] However, the current traditional scheduling method has significant limitations. The priority setting lacks dynamic adaptability and cannot adjust the priority sequence according to the real-time state of the task (such as changes in the deadline and fluctuations in the data processing level) and the dynamic changes in the business demand of the industrial control scene, which may result in that the key task cannot be executed in priority due to unreasonable initial priority setting. The resource allocation does not consider the quantitative evaluation of the real-time resource state, making it difficult to accurately select potential available resources and easily leading to overallocation or underallocation of resources, thereby reducing resource utilization. The traditional scheduling method lacks the ability to respond to sudden scenarios. When facing sudden situations such as resource overload, task emergency insertion, or temporary resource failure, the traditional scheduling method cannot quickly deduce the adjustment process of resource allocation and scheduling path, which may easily lead to task blocking, execution delay, and even system crash. No effective compensation optimization mechanism is established. In the event of a sudden scenario, the affected tasks cannot obtain reasonable resource compensation according to the priority, which affects the overall scheduling efficiency and reliability of the system. SUMMARY

[0004] Therefore, it is necessary to provide a dynamic-priority-based integrated SOC industrial computer real-time scheduling method, system, device and medium to solve the above technical problems.

[0005] In a first aspect, the application provides a dynamic-priority-based integrated SOC industrial computer real-time scheduling method, comprising:

[0006] Collecting system state data of the integrated SOC industrial computer, the system state data including task state data and resource state data;

[0007] Based on the resource state data, a potential available resource set is determined by a state quantification method;

[0008] Based on the task state data, the priority of each task is evaluated based on a dynamic priority evaluation rule preset in combination with the business demand of the industrial control scene, and a task priority sequence is generated;

[0009] Based on the potential available resource set and the task priority sequence, a preliminary resource allocation-scheduling path scheme is generated by using a pre-trained reinforcement learning algorithm according to a preset state space and action space;

[0010] Based on the preliminary resource allocation-scheduling path scheme, a simulation result of the burst scene is obtained by simulating the resource supply-demand fluctuation and task priority mutation in the preset burst scene and deducing the adjustment process of the resource allocation and scheduling path;

[0011] According to the simulation result of the burst scene, an optimized resource allocation-scheduling path scheme is obtained by using a priority-based compensation strategy to optimize the preliminary resource allocation-scheduling path scheme;

[0012] According to the optimized resource allocation-scheduling path scheme, an integrated soc industrial computer scheduling instruction is generated, which is used to drive the integrated soc industrial computer to execute tasks according to the optimized resource allocation logic and scheduling path.

[0013] In one embodiment, based on the resource state data, a potential available resource set is determined by a state quantification method, including:

[0014] The resource state data is classified according to resource types to obtain a resource classification result, and the resource types include computing resources, storage resources, and communication resources;

[0015] For each type of resource in the resource classification result, a state quantification index is configured based on the hardware configuration parameters and the regular running load of the integrated soc industrial computer;

[0016] Based on the state quantification index, the resource state data of each type of resource is quantified to obtain quantified values of each type of resource;

[0017] Based on the quantified values of each type of resource, resources whose quantified values meet the corresponding preset available resource threshold are selected to form a potential available resource set.

[0018] In one embodiment, the dynamic priority evaluation rule includes weight configurations of key parameters of each task in different industrial control business scenes;

[0019] Based on the task state data, the priority of each task is evaluated based on a dynamic priority evaluation rule preset in combination with the business demand of the industrial control scene, and a task priority sequence is generated, including:

[0020] From the task state data, the key parameters associated with the task in the industrial control scene are extracted, and the key parameters include task deadline, data processing magnitude, associated equipment dependency, and task execution urgency identifier;

[0021] Based on the dynamic priority evaluation rule, the key parameters of each task are weighted calculated with the corresponding weight to obtain the priority quantitative score of each task;

[0022] The priority quantitative scores of all tasks are sorted in descending order, and a task priority sequence is generated according to the sorting result.

[0023] In one embodiment, in the pre-training reinforcement learning algorithm, the state space is composed of the feature representation of the set of potential available resources and the feature representation of the task priority sequence, and the action space is the set of feasible resource allocation modes and scheduling path types;

[0024] Based on the set of potential available resources and the task priority sequence, a pre-training reinforcement learning algorithm is used to generate a preliminary resource allocation-scheduling path scheme according to the preset state space and action space, including:

[0025] The set of potential available resources and the task priority sequence are input into the state space for feature encoding to form initial state data;

[0026] Based on the initial state data, the corresponding resource allocation mode and scheduling path type are matched in the action space to obtain an action output result;

[0027] An adaptive adjustment factor is used to evaluate the adaptability of the initial state data and the action output result, and an adaptability evaluation result is generated;

[0028] Based on the action output result and the adaptability evaluation result, combined with the resource capacity of the set of potential available resources and the task priority order of the task priority sequence, a preliminary resource allocation-scheduling path scheme is generated, which includes the resource type allocated to each task, the resource allocation amount of each resource type, and the node order information of the scheduling execution.

[0029] In one embodiment, based on the action output result and the adaptability evaluation result, combined with the resource capacity of the set of potential available resources and the task priority order of the task priority sequence, a preliminary resource allocation-scheduling path scheme is generated, using the following formula:

[0030]

[0031] Wherein, is the probability of allocating resources to a given task under the condition that is a nonlinear transformation with an index of is a priority of the task , is a nonlinear transformation with an index of is an adaptability evaluation result of the task to the resource , is a nonlinear transformation with an index of is a probability of resource conflict between the task and the resource , is a nonlinear transformation with an index of is a probability of resource conflict between the task and the resource , , and are priority, adaptability and potential available resource adaptability adjustment factors respectively, is a strategy temperature parameter, is a minimum constant to avoid zero division, is an exponential function, is a set of all potential optional resources, and k is a resource index, is a time-dependent decay function, where t is the current system time, is a resource suggested execution time window assigned to the task .

[0032] In one of the embodiments, the preset burst scenarios include a resource overload scenario, a task emergency insertion scenario and a resource temporary failure scenario;

[0033] Based on the preliminary resource allocation-scheduling path scheme, by simulating the resource supply and demand fluctuation and the task priority mutation in the preset burst scenario, the adjustment process of the resource allocation and scheduling path is deduced, and the burst scenario simulation result is obtained, including:

[0034] For each preset burst scenario type, resource supply and demand fluctuation parameters and task priority mutation parameters in the corresponding scenario are generated, wherein the resource supply and demand fluctuation parameters include resource supply amount change parameters, resource demand amount change parameters in the corresponding scenario, and the task priority mutation parameters include new task priority parameters and original task priority adjustment parameters in the corresponding scenario;

[0035] Based on the preliminary resource allocation-scheduling path scheme, the resource supply and demand fluctuation data are matched with the resource type and the resource allocation amount corresponding to the resource type to determine the target resource;

[0036] Based on the preliminary resource allocation-scheduling path scheme, the task priority mutation parameters are matched with the node order information of the scheduling execution to determine the target task;

[0037] For each preset burst scenario type, based on the resource supply and demand fluctuation parameter, the allocation amount of the target resource is deduced to increase or decrease, and the target resource allocation amount adjustment scheme is obtained;

[0038] For each preset burst scenario type, based on the task priority mutation parameter, the scheduling node insertion, deletion or ordering re-adjustment process of the target task is deduced, and the target task scheduling node change scheme is obtained;

[0039] Collect the adjustment record of the target resource allocation amount under each preset burst scenario type, the target task scheduling node change record, and the quantitative score comparison data before and after the task priority mutation, and form the burst scenario simulation result.

[0040] In one of the embodiments, according to the burst scenario simulation result, a priority-based compensation strategy is used to optimize the preliminary resource allocation-scheduling path scheme, and an optimized resource allocation-scheduling path scheme is obtained, including:

[0041] Based on the burst scenario simulation result, the target task affected by the burst factor in the preliminary resource allocation-scheduling path scheme is identified;

[0042] According to the task priority sequence, the compensation resource allocation order of the target task is determined;

[0043] According to the compensation resource allocation order, the resource type allocated to the target task, the resource allocation amount of each resource type, and the node order of scheduling execution are adjusted to form an optimized resource allocation-scheduling path scheme.

[0044] In a second aspect, the application also provides an integrated soc industrial computer real-time scheduling system based on dynamic priority, including:

[0045] A multi-source state acquisition module is used to acquire system state data of the integrated soc industrial computer, and the system state data includes task state data and resource state data;

[0046] An available resource evaluation module is used to determine a potential available resource set based on the resource state data through a state quantization method;

[0047] A priority evaluation module is used to evaluate the priority of each task based on the task state data and in combination with a dynamic priority evaluation rule preset according to the business demand of the industrial scene, and a task priority sequence is generated;

[0048] A scheduling scheme generation module is used to generate a preliminary resource allocation-scheduling path scheme based on the potential available resource set and the task priority sequence, and in combination with a pre-trained reinforcement learning algorithm and a preset state space and action space;

[0049] The deduction simulation module is configured to deduce an adjustment process of the resource allocation and the scheduling path based on the preliminary resource allocation-scheduling path scheme by simulating resource supply-demand fluctuation and task priority mutation under preset burst scenarios to obtain a burst scenario simulation result.

[0050] The compensation optimization module is configured to optimize the preliminary resource allocation-scheduling path scheme based on the burst scenario simulation result by using a priority-based compensation strategy to obtain an optimized resource allocation-scheduling path scheme.

[0051] The scheduling instruction generation module is configured to generate an integrated SOC industrial computer scheduling instruction based on the optimized resource allocation-scheduling path scheme, and the integrated SOC industrial computer scheduling instruction is configured to drive the integrated SOC industrial computer to execute tasks according to the optimized resource allocation logic and the scheduling path.

[0052] In a third aspect, the present application further provides a computer device including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the real-time scheduling method of the integrated SOC industrial computer based on dynamic priority when executing the computer program.

[0053] In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, and the computer program implements the steps of the real-time scheduling method of the integrated SOC industrial computer based on dynamic priority when executed by a processor.

[0054] The real-time scheduling method, system, device and medium of the integrated SOC industrial computer based on dynamic priority improve resource utilization and dynamic adaptability by collecting system state data of the integrated SOC industrial computer and determining a potential available resource set based on resource state data using a state quantization method, overcome the rigidity problem of traditional fixed priority setting by generating a task priority sequence in combination with task state data and dynamic priority evaluation rules, enhance the adaptive optimization capability of the scheduling process by generating a preliminary resource allocation-scheduling path scheme based on a state space and an action space using a pre-trained reinforcement learning algorithm, improve the resilience of the system in dealing with burst situations such as resource overload and task urgent insertion by simulating preset burst scenarios to deduce the adjustment process of the resource allocation and the scheduling path, and optimizing the preliminary scheme in combination with a priority-based compensation strategy, reduce task blocking and delay, and generate integrated SOC industrial computer scheduling instructions to drive execution and guarantee the reliable landing of the optimized scheme. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiments or the related art description will be briefly introduced. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor.

[0056] Figure 1 The flow chart of the real-time scheduling method of the integrated soc industrial computer based on dynamic priority of the present application;

[0057] Figure 2 The system structure diagram of the real-time scheduling system of the integrated soc industrial computer based on dynamic priority of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0059] In one embodiment, as shown in Figure 1 , a real-time scheduling method of an integrated soc industrial computer based on dynamic priority is provided. This embodiment takes the method applied to a terminal as an example. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server.

[0060] The hardware architecture of the present embodiment takes an integrated soc industrial computer as a terminal (such as using a SOC chip with model XXX, integrating 4-core ARM Cortex-A53 computing core, 1GB on-chip DDR4 storage), which is connected to multiple types of industrial sensors, actuators, local storage modules and industrial Ethernet communication modules. The server communicates with the integrated soc industrial computer through the industrial network, and is used for pre-training the reinforcement learning model and storing the scene configuration data. The application scenarios include: when there are multiple concurrent task execution, resource dynamic fluctuation and burst scenario (such as emergency task insertion, resource failure) requirements in industrial production, the sensor collects the task state and resource state data of the integrated soc industrial computer and transmits them to the integrated soc industrial computer. The integrated soc industrial computer calls the present method to generate optimized scheduling instructions, which are sent to the actuators through the communication module to execute tasks, and the server synchronously provides model parameter support and scene simulation data.

[0061] In the present embodiment, the method includes the following steps:

[0062] S01, collecting system state data of the integrated soc industrial computer, the system state data including task state data and resource state data.

[0063] The system state data is task state data (such as task identification, execution progress, deadline, and associated device dependency) and resource state data (such as computing resources of an integrated SOC module, storage resources, and communication resources of an industrial computer), which can be obtained by using the hardware monitoring sensor, operating system level task scheduling interface, industrial network communication protocol, or external data acquisition module built in the integrated SOC industrial computer to provide comprehensive and dynamic basic data input.

[0064] S02, based on the resource state data, determining a set of potential available resources by a state quantization method.

[0065] The state quantization method refers to a technical means for converting resource state into a comparable quantized value through mathematical or logical processing, and the set of potential available resources is a resource group that meets the preset conditions after quantization and evaluation; in implementation, the resource state data can be classified by type, and based on the hardware configuration and running load configuration of the integrated SOC industrial computer, quantization indicators (such as threshold or scoring rules) are configured, then the resources of each type are quantized (such as numerical normalization, weighted calculation, or probability model application), and the quantization results are compared with the preset available resource threshold to dynamically identify available resources, adapt to resource fluctuations, and improve resource utilization efficiency.

[0066] S03, based on the task state data, evaluating the priority of each task according to the dynamic priority evaluation rule preset according to the business demand of the industrial control scene, and generating a task priority sequence.

[0067] The dynamic priority evaluation rule is an evaluation logic preset according to different industrial control business scenes (such as production control, device monitoring, or emergency processing), which can include weight configuration, scoring model, or adaptive adjustment mechanism of key parameters (such as deadline or dependency), and the task priority sequence is a task list arranged in order of priority (descending or ascending) after evaluation; in implementation, the key parameters can be extracted from the task state data, the priority quantization score of each task can be calculated based on the dynamic priority evaluation rule (such as weighted summation, fuzzy logic, or machine learning model), and all scores can be sorted to generate a sequence, which dynamically adapts to changes in the industrial control scene and ensures that high-priority tasks are scheduled first.

[0068] S04, based on the set of potential available resources and the task priority sequence, using a pre-trained reinforcement learning algorithm to generate a preliminary resource allocation-scheduling path scheme according to the preset state space and action space.

[0069] Wherein, the state space is a multi-dimensional decision environment model composed of the characteristics of the set of potential available resources (such as resource type, capacity, real-time load feature vector or embedding representation) and the feature representation of the task priority sequence (such as task priority score, sequence encoding of dependency relationship); the action space is the set of feasible resource allocation methods (such as exclusive allocation, shared allocation) and scheduling path types (such as sequential execution, parallel execution); in implementation, the resource set and task sequence can be input into the state space for feature coding to form the initial state data, and the pre-trained reinforcement learning model (such as Q-learning or policy gradient-based algorithm) is used to match the corresponding resource allocation and scheduling path action in the action space, and the resource adaptation adjustment factor is used to adaptively evaluate the action output, and the resource capacity and priority order are combined to generate a preliminary scheme containing resource allocation type, allocation amount and node order information, so that the scheduling decision has adaptive optimization capability.

[0070] S05, based on the preliminary resource allocation-scheduling path scheme, simulates the resource supply and demand fluctuation and task priority mutation under the preset burst scenario, deduces the adjustment process of resource allocation and scheduling path, and obtains the burst scenario simulation result.

[0071] Wherein, the preset burst scenario is an abnormal working condition that may affect the stability of the system, including resource overload, task emergency insertion and temporary resource failure; resource supply and demand fluctuation refers to the dynamic change of resource supply or demand; task priority mutation includes the addition of high priority tasks or the emergency adjustment of original task priority; in implementation, different burst scenario types (such as resource overload scenario, SOC computing core utilization rate increases to 90%) can be used to generate corresponding resource fluctuation parameters and priority mutation parameters, which are matched with the resource type, allocation amount and node order information in the preliminary scheme to determine the target resources and target tasks, and the increase and decrease process of the target resource allocation amount (such as adjusting the allocation amount of SOC computing core from 1 core to 2 cores) and the insertion, deletion or re-adjustment process of the target task scheduling node are deduced through simulation, and the quantitative simulation results are formed by collecting resource adjustment records, node change records and priority comparison data, which provide decision basis for subsequent compensation optimization.

[0072] S06, according to the burst scenario simulation result, a priority-based compensation strategy is adopted to optimize the preliminary resource allocation-scheduling path scheme, and an optimized resource allocation-scheduling path scheme is obtained.

[0073] The priority-based compensation strategy is a mechanism for compensating tasks affected by sudden factors according to a task priority sequence, including resource reallocation, scheduling sequence adjustment or resource reservation strategy; in implementation, the target task affected by resource fluctuation or priority mutation in the preliminary scheme can be identified, the compensation sequence can be determined according to the task priority sequence, and the target task can be compensated and optimized by dynamically reallocating resource types and resource amounts, adjusting the execution sequence of the scheduling node or introducing redundant resources, to form an optimized scheme that can adapt to sudden scenarios and guarantee the execution of high-priority tasks.

[0074] S07, according to the optimized resource allocation-scheduling path scheme, generating integrated soc industrial computer scheduling instructions, the integrated soc industrial computer scheduling instructions are used to drive the integrated soc industrial computer to execute tasks according to the optimized resource allocation logic and scheduling path.

[0075] The integrated soc industrial computer scheduling instruction is a set of control commands that can be parsed and executed by the integrated soc industrial computer, covering resource allocation logic (such as allocation rules for computing, storage or communication resources) and scheduling path (such as task execution sequence or parallel processing instructions); in implementation, the optimized scheme can be converted into an instruction format (such as a data packet or API call based on an industrial communication protocol) that can be recognized by the integrated soc industrial computer, the resource allocation details and scheduling sequence can be encoded into specific operation commands, and the integrated soc industrial computer hardware interface (such as PLC or industrial bus) can be combined to issue and execute, ensuring that the optimized scheme can be reliably implemented.

[0076] The above-mentioned real-time scheduling method of integrated soc industrial computer based on dynamic priority, by collecting system state data of integrated soc industrial computer, dynamically determines the potential available resource set based on resource state data using state quantization method, realizes real-time quantization evaluation and screening of computing, storage and communication resources; combined with the dynamic priority evaluation rule preset by the task state data and the industrial scene business demand, the priority of each task is evaluated to generate a task priority sequence, solving the limitation that the fixed priority setting cannot execute the key task preferentially; using pre-trained reinforcement learning algorithm, according to the preset state space and action space, a preliminary resource allocation-scheduling path scheme is generated, and the adjustment process of resource allocation and scheduling path is deduced by simulating the preset sudden scene, and the preliminary scheme is optimized by combining the priority-based compensation strategy, to generate integrated soc industrial computer scheduling instructions to drive execution, improving the real-time performance and self-adaptive ability of the system to cope with resource supply and demand fluctuations and task priority mutation, reducing task blocking and delay, and guaranteeing the priority execution of key tasks.

[0077] In one embodiment, based on resource state data, the potential available resource set is determined by a state quantization method, including:

[0078] S11, classify the resource state data according to resource types to obtain resource classification results, the resource types including computing resources, storage resources, and communication resources;

[0079] S12, for each type of resource in the resource classification results, configure state quantization indicators based on integrated SOC industrial computer hardware configuration parameters and regular running loads;

[0080] S13, based on the state quantization indicators, quantize the resource state data of each type of resource to obtain quantized values of each type of resource;

[0081] S14, based on the quantized values of each type of resource, filter out resources whose quantized values meet corresponding preset available resource thresholds to form a set of potential available resources.

[0082] Specifically, the collected resource state data can be classified according to resource types, the resource types including computing resources (including computing core resources of integrated SOC modules), storage resources (including on-chip storage resources of integrated SOC modules), and communication resources, to obtain resource classification results; for each type of resource after classification, state quantization indicators are configured based on integrated SOC industrial computer hardware configuration parameters (such as SOC module model, core number, and on-chip storage capacity) and regular running loads (such as setting the average utilization rate threshold of SOC computing core to 60%, the on-chip storage occupancy threshold to 70%, and the communication resource indicator to bandwidth idle rate); based on these state quantization indicators, real-time state data of each type of resource is quantized, such as converting original data into comparable quantized values through a normalization algorithm or a weighted scoring model; the quantized values of each type of resource are compared with preset available resource thresholds (such as setting the computing resource threshold to an available rate greater than 30%, and the specific value can be set according to historical records in actual situations), and resources meeting the conditions are filtered out to form a set of potential available resources. Through dynamic quantization and threshold filtering, the real-time and accuracy of resource evaluation are ensured, and the adaptive ability of integrated SOC industrial computer resource scheduling is improved.

[0083] In one of the embodiments, the dynamic priority evaluation rule includes weight configurations of each task critical parameter in different industrial control business scenarios;

[0084] Based on the task state data, the dynamic priority evaluation rule preset in combination with the industrial control scenario business demand is used to evaluate the priority of each task to generate a task priority sequence, including:

[0085] S21, from the task state data, extract task-associated critical parameters in the industrial control scenario, the critical parameters including task deadline, data processing level, associated device dependency, and task execution urgency identifier;

[0086] S22, based on the dynamic priority evaluation rule, the key parameters of each task are weighted calculated with the corresponding weight to obtain the priority quantitative score of each task;

[0087] S23, the priority quantitative scores of all tasks are sorted in descending order, and a task priority sequence is generated according to the sorting result.

[0088] Exemplarily, based on the task state data, the key parameters associated with the task in the industrial control scene can be extracted by analyzing the task description or real-time monitoring interface, such as task deadline (such as remaining time to deadline), data processing level (such as data size to be processed), associated device dependency (such as the number or criticality of dependent devices), and task execution urgency identifier (such as emergency flag); combined with the dynamic priority evaluation rule preset according to the business demand of the industrial control scene (including the weight configuration of each key parameter corresponding to different business scenarios, for example, in the production control scene, the weight of task deadline is higher), the extracted key parameters are weighted calculated with the corresponding weight (such as using linear weighted sum formula:

[0089]

[0090] Among them, 、 、 and The weight value can be dynamically adjusted according to the scene), to obtain the priority quantitative score of each task; the quantitative scores of all tasks are sorted in descending order (such as quicksort or heapsort), and a task priority sequence is generated according to the sorting result, to ensure that high-priority tasks are scheduled first. Through the consecutive operations of parameter extraction, weighted calculation and sorting, the dynamic quantitative evaluation of task priority is realized, and the real-time performance and business adaptability of integrated soc industrial computer scheduling are improved.

[0091] In one of the embodiments, in the pre-trained reinforcement learning algorithm, the state space is composed of the feature representation of the set of potentially available resources and the feature representation of the task priority sequence, and the action space is the set of feasible resource allocation methods and scheduling path types;

[0092] Based on the set of potentially available resources and the task priority sequence, a pre-trained reinforcement learning algorithm is used to generate a preliminary resource allocation-scheduling path scheme according to the preset state space and action space, including:

[0093] S31, input the set of potentially available resources and the task priority sequence into the state space for feature encoding to form initial state data;

[0094] S32, based on the initial state data, match the corresponding resource allocation method and scheduling path type in the action space to obtain the action output result;

[0095] S33, using a resource adaptation adjustment factor, adaptively evaluating the initial state data and the action output result to generate an adaptability evaluation result;

[0096] S34, based on the action output result and the adaptability evaluation result, combining the resource capacity of the set of potentially available resources and the task priority order of the task priority sequence to generate a preliminary resource allocation-scheduling path scheme, which contains the resource type allocated to each task, the resource allocation amount of each resource type, and the node order information of the scheduling execution.

[0097] Specifically, the set of potentially available resources and the task priority sequence can be input into the state space for feature encoding, such as using a pre-trained feature extraction network (full connection layer or attention mechanism) to fuse the resource features and the task sequence to form initial state data; based on the initial state data, the corresponding resource allocation mode and scheduling path type are matched in the action space through a policy network (such as a model based on Q-learning or policy gradient), to obtain an action output result; a resource adaptation adjustment factor (such as a weight parameter and is used to adjust the influence of priority, adaptability and resource capacity) is used to adaptively evaluate the initial state data and the action output result, and an adaptability evaluation result is generated by calculating the adaptability score (such as using a similarity measure or a weighting function); based on the action output result and the adaptability evaluation result, the resource capacity (such as the number of remaining cores) of the set of potentially available resources and the task priority order of the task priority sequence are combined, and an optimization algorithm (such as a greedy strategy or a probability allocation model) is used to generate a preliminary resource allocation-scheduling path scheme, which contains the resource type (such as computing, storage or communication resources) allocated to each task, the resource allocation amount of each resource type, and the node order information (such as the time sequence or dependency relationship of task execution) of the scheduling execution.

[0098] In one of the embodiments, S41, based on the action output result and the adaptability evaluation result, combining the resource capacity of the set of potentially available resources and the task priority order of the task priority sequence to generate a preliminary resource allocation-scheduling path scheme, using the following formula:

[0099]

[0100] wherein, is the probability of allocating resource to the given task , is a nonlinear transformation with an exponential of applied to the priority of task , denotes the task With resources Adaptability assessment results The applied index is Nonlinear transformation, To assess potential available resources The applied index is Nonlinear transformation, For the task With resources The probability of resource conflicts between them , and These are the priority, adaptability, and potential availability resource adaptation adjustment factors, respectively. For strategy temperature parameters, To avoid the smallest constant that divides by zero, It is an exponential function. Let k be the set of all potential available resources, and k be the resource index. Let be a time-dependent decay function, where t is the current system time. For resources Assigned to task The recommended execution time window.

[0101] For example, the formula calculates the value of a given task based on the action output (such as the matching result of resource allocation method and scheduling path type) and the adaptability evaluation result (such as task and resource adaptability score), combined with the resource capacity of the potentially available resource set (such as the number of remaining cores or on-chip storage space) and the task priority order of the task priority sequence. Resource allocation under certain conditions probabilistic allocation You can first work on the task. priority The applied index is Nonlinear transformations (such as) (to amplify the impact of high-priority tasks), on tasks With resources Adaptability assessment results The applied index is Nonlinear transformations (such as) To enhance adaptability differences, for potential available resources The applied index is Nonlinear transformations (such as) To adjust resource capacity weights and introduce resource conflict probability. (Based on historical data or real-time monitoring probability values) to reduce conflict risk; through exponential functions and strategy temperature parameters ( scaling the transformed values and using a small constant e (such as ) to avoid division by zero to ensure denominator stability; the numerator part calculates the combined effect of priority, fitness, and resource capacity, and the denominator normalizes by summing over all potential alternative resource sets M (such as resource index k iterating over all available resources) to form a softmax-like probability distribution; multiplying by resource capacity and applying a time-dependent decay function (such as an exponential decay function, where t is the current system time, is the proposed execution time window) to dynamically adjust the allocation amount over time; the calculated probabilistic allocation amount is used to determine the resource type assigned to each task, the resource allocation amount of each resource type, and the node order information for scheduling execution, thereby generating an implementable preliminary scheme.

[0102] In one embodiment, the preset burst scenarios include a resource overload scenario, a task emergency insertion scenario, and a resource temporary failure scenario;

[0103] Based on the preliminary resource allocation-scheduling path scheme, by simulating the resource supply and demand fluctuations and task priority mutations in the preset burst scenarios, the adjustment process of the resource allocation and scheduling path is deduced, and the burst scenario simulation results are obtained, including:

[0104] S51, for each preset burst scenario type, generate resource supply and demand fluctuation parameters and task priority mutation parameters in the corresponding scenario, wherein the resource supply and demand fluctuation parameters include resource supply amount change parameters, resource demand amount change parameters in the corresponding scenario, and the task priority mutation parameters include new task priority parameters and original task priority adjustment parameters in the corresponding scenario;

[0105] S52, based on the preliminary resource allocation-scheduling path scheme, match the resource supply and demand fluctuation data with the resource type and the resource allocation amount corresponding to the resource type to determine the target resource;

[0106] S53, based on the preliminary resource allocation-scheduling path scheme, match the task priority mutation parameters with the node order information for scheduling execution to determine the target task;

[0107] S54, for each preset burst scenario type, based on the resource supply and demand fluctuation parameters, deduce the allocation amount increase or decrease process of the target resource, and obtain the target resource allocation amount adjustment scheme;

[0108] S55, for each preset burst scenario type, based on the task priority mutation parameters, deduce the scheduling node insertion, deletion, or ordering re-adjustment process of the target task, and obtain the target task scheduling node change scheme;

[0109] S56, collect the adjustment record of the target resource allocation amount, the target task scheduling node change record and the quantitative score comparison data before and after the task priority mutation under each preset burst scenario type, and form a burst scenario simulation result.

[0110] Specifically, for preset burst scenario types (including resource overload scenario, task emergency insertion scenario and resource temporary failure scenario), resource supply and demand fluctuation parameters (such as resource supply amount change parameter can be set to 20% decrease of available computing resources, resource demand amount change parameter can be set to 30% increase of storage resource request amount) and task priority mutation parameters (such as new task priority parameter can be set to high priority score 90, original task priority adjustment parameter can be dynamically improved weight based on emergency flag) under corresponding scenarios can be generated by a scenario simulator; based on the preliminary resource allocation-scheduling path scheme, the resource supply and demand fluctuation data are matched with the resource types in the scheme and the allocation amount of each type of resource (such as using key-value mapping or rule engine lookup), the target resource affected by the fluctuation is determined (such as identifying that the computing resource A is the target resource under the overload scenario), and the task priority mutation parameters are matched with the node sequence information (such as task execution time sequence) of the scheduling execution (such as priority threshold comparison or dependency analysis), the target task (such as the emergency insertion task For each preset burst scenario type, the allocation amount increase or decrease process of the target resource is deduced based on the resource supply and demand fluctuation parameters (such as using linear adjustment algorithm to simulate dynamic scaling of resource allocation amount, and obtaining the target resource allocation amount adjustment scheme), and the scheduling node insertion (such as inserting a new node in the sequence using linked list operation), deletion (such as removing low priority tasks) or ordering re-adjustment process (such as applying bubble sort to rearrange according to new priority) of the target task is deduced based on the task priority mutation parameters, and the target task scheduling node change scheme is obtained; the adjustment record of the target resource allocation amount (such as resource allocation amount change value), the target task scheduling node change record (such as node insertion position index) and the quantitative score comparison data before and after the task priority mutation (such as the score difference before and after the mutation) under each scenario type can be collected, and these data are integrated to form a structured burst scenario simulation result, which provides a quantitative basis for subsequent compensation optimization, and improves the resilience of the system to deal with burst scenarios.

[0111] In one of the embodiments, according to the burst scenario simulation result, a priority-based compensation strategy is adopted to optimize the preliminary resource allocation-scheduling path scheme, and an optimized resource allocation-scheduling path scheme is obtained, including:

[0112] S61, based on the burst scenario simulation result, identifying the target task affected by the burst factor in the preliminary resource allocation-scheduling path scheme;

[0113] S62, determine the compensation resource allocation order of the target task according to the task priority sequence;

[0114] S63, adjust the resource type allocated to the target task, the resource allocation amount of each resource type, and the node order of scheduling execution according to the compensation resource allocation order, and form an optimized resource allocation-scheduling path scheme.

[0115] Illustratively, the target task affected by the burst factor (such as resource overload or task urgent insertion) in the preliminary resource allocation-scheduling path scheme can be identified based on the burst scenario simulation result through a data matching algorithm (such as key field query or rule engine filtering), such as identifying the task delayed due to resource shortage ); according to the task priority sequence (such as the sequence arranged in descending order of quantified scores), the priority queue management mechanism (such as the minimum heap or priority queue data structure) is used to determine the compensation resource allocation order of the target task (that is, the high-to-low priority is processed in turn); according to the compensation resource allocation order, the resource type allocated to the target task is dynamically adjusted (such as replacing the computing resource with the standby communication resource through the resource reallocation logic), the resource allocation amount of each resource type (such as increasing the resource quota of the high-priority task using the linear interpolation or scaling algorithm), and the node order of scheduling execution (such as rearranging the task execution sequence by applying the insertion sort or dependency adjustment algorithm), forming an optimized resource allocation-scheduling path scheme, ensuring that high-priority tasks still have resource guarantees under burst scenarios through compensation strategies, and improving the overall scheduling efficiency and reliability of the system.

[0116] The above-mentioned integrated SOC industrial computer real-time scheduling method based on dynamic priority, by collecting system state data of the integrated SOC industrial computer, including task state data and resource state data, and based on the resource state data, using a state quantization method to determine a set of potentially available resources, including classifying resource data by type, configuring state quantization indicators based on hardware parameters and regular running load, performing quantization processing to obtain resource quantization values of each type, and filtering resources that meet the preset available resource threshold, dynamically identifying available resources, improving the accuracy and adaptability of resource utilization, and overcoming the limitations of traditional resource allocation surplus or deficiency. Combined with the task state data and the dynamic priority evaluation rules (including the weight configuration of task critical parameters such as deadline, data processing level, associated device dependency, and task execution urgency identification in different scenarios) preset by the industrial scene business demand, the priority of each task is evaluated, the priority quantization score is obtained by weighted calculation through the extraction of key parameters, and the task priority sequence is generated in descending order, realizing dynamic adjustment of priority, ensuring high-priority tasks to be scheduled first, and solving the problem of fixed priority rigidity. A pre-trained reinforcement learning algorithm is used to generate a preliminary resource allocation-scheduling path scheme based on a preset state space (composed of feature representations of the set of potentially available resources and feature representations of the task priority sequence) and an action space (a set of feasible resource allocation methods and scheduling path types). The initial state data is formed by feature encoding, the action output result is matched, and a resource adaptation adjustment factor (such as ) is used for adaptability evaluation, and the formula

[0117]

[0118] is used for probabilistic allocation to enhance the adaptive optimization capability of scheduling decision. By simulating the resource supply and demand fluctuations and task priority mutations under preset sudden scenarios (such as resource overload, task emergency insertion, and temporary resource failure), the adjustment process of resource allocation and scheduling path is deduced, including generating resource supply and demand fluctuation parameters and task priority mutation parameters, matching target resources and target tasks, deducing target resource allocation adjustment scheme and target task scheduling node change scheme, collecting adjustment records and comparison data to form sudden scenario simulation results, and combining the priority-based compensation strategy to optimize the preliminary scheme, such as identifying target tasks affected by sudden factors, determining the compensation order according to the task priority sequence, adjusting the resource type, allocation amount, and node order, improving the resilience of the system to resource fluctuations and priority mutations, and reducing task blocking and delay. The integrated SOC industrial computer scheduling instruction is generated to drive the integrated SOC industrial computer to execute, ensuring the landing of the scheme. Through this method, the real-time performance, adaptive capability, and sudden scenario response level of the integrated SOC industrial computer resource scheduling can be improved, ensuring the priority execution of critical tasks and improving the overall efficiency and reliability.

[0119] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0120] Based on the same inventive concept, the embodiments of the present application also provide a dynamic priority-based integrated soc industrial computer real-time scheduling system for implementing the above-mentioned dynamic priority-based integrated soc industrial computer real-time scheduling method. The problem-solving implementation scheme provided by the system is similar to the implementation scheme described in the above method, so the specific limitations in one or more dynamic priority-based integrated soc industrial computer real-time scheduling system embodiments provided below can refer to the limitations of the dynamic priority-based integrated soc industrial computer real-time scheduling method described above, which will not be repeated here.

[0121] In one exemplary embodiment, as shown in Figure 2 a dynamic priority-based integrated soc industrial computer real-time scheduling system is provided, comprising:

[0122] The multi-source state acquisition module 101 is configured to acquire system state data of the integrated soc industrial computer, and the system state data includes task state data and resource state data.

[0123] The available resource evaluation module 102 is configured to determine a set of potentially available resources based on the resource state data by a state quantization method.

[0124] The priority evaluation module 103 is configured to evaluate the priority of each task based on the task state data and in combination with a dynamic priority evaluation rule preset according to the business requirements of the industrial scene, and generate a task priority sequence.

[0125] The scheduling scheme generation module 104 is configured to generate a preliminary resource allocation-scheduling path scheme based on the set of potentially available resources and the task priority sequence, and by using a pre-trained reinforcement learning algorithm according to a preset state space and action space.

[0126] The deduction simulation module 105 is configured to deduce an adjustment process of the resource allocation and scheduling path based on the preliminary resource allocation-scheduling path scheme by simulating resource supply-demand fluctuations and task priority mutation conditions under preset burst scenarios, and obtain a burst scenario simulation result.

[0127] The compensation optimization module 106 is configured to optimize the preliminary resource allocation-scheduling path scheme according to the burst scenario simulation result by using a priority-based compensation strategy, and obtain an optimized resource allocation-scheduling path scheme.

[0128] The scheduling instruction generation module 107 is configured to generate an integrated soc industrial computer scheduling instruction according to the optimized resource allocation-scheduling path scheme, and the integrated soc industrial computer scheduling instruction is configured to drive the integrated soc industrial computer to execute tasks according to the optimized resource allocation logic and scheduling path.

[0129] In one of the embodiments, the available resource evaluation module 102 is further configured to:

[0130] The resource state data is classified according to resource types to obtain a resource classification result, and the resource types include computing resources, storage resources, and communication resources.

[0131] For each type of resource in the resource classification result, a state quantization index is configured based on the hardware configuration parameters of the integrated soc industrial computer and the regular running load.

[0132] Based on the state quantization index, the resource state data of each type of resource is quantized to obtain quantized values of each type of resource.

[0133] Based on the quantized values of each type of resource, resources whose quantized values meet corresponding preset available resource thresholds are screened out to form a set of potential available resources.

[0134] In one of the embodiments, the dynamic priority evaluation rule in the priority evaluation module 103 includes weight configurations of critical parameters of each task in different industrial control business scenarios.

[0135] Based on the task state data, the dynamic priority evaluation rule preset according to the business demand of the industrial control scenario is used to evaluate the priority of each task to generate a task priority sequence, including:

[0136] From the task state data, critical parameters associated with the task in the industrial control scenario are extracted, and the critical parameters include task deadline, data processing level, associated equipment dependency, and task execution urgency identifier.

[0137] Based on the dynamic priority evaluation rule, the critical parameters of each task are weighted calculated with corresponding weights to obtain a priority quantization score of each task.

[0138] All tasks are sorted in descending order of their priority quantification scores, and a task priority sequence is generated based on the sorting results.

[0139] In one embodiment, in the pre-trained reinforcement learning algorithm of the scheduling scheme generation module 104, the state space is composed of the feature representation of the set of potential available resources and the feature representation of the task priority sequence, and the action space is a set of feasible resource allocation methods and scheduling path types.

[0140] Based on the set of potentially available resources and the sequence of task priorities, a pre-trained reinforcement learning algorithm is used to generate a preliminary resource allocation-scheduling path scheme according to a preset state space and action space, including:

[0141] The set of potential available resources and the sequence of task priorities are input into the state space for feature encoding to form initial state data;

[0142] Based on the initial state data, the corresponding resource allocation method and scheduling path type are matched in the action space to obtain the action output result;

[0143] A resource adaptation adjustment factor is used to evaluate the compatibility between the initial state data and the action output results, and an adaptation evaluation result is generated.

[0144] Based on the action output results and the adaptability evaluation results, combined with the resource capacity of the potential available resource set and the task priority order of the task priority sequence, a preliminary resource allocation-scheduling path scheme is generated. The preliminary resource allocation-scheduling path scheme includes the resource type allocated to each task, the resource allocation amount of each resource type, and the node sequence information for scheduling execution.

[0145] In one embodiment, the scheduling scheme generation module 104 is further configured to generate a preliminary resource allocation-scheduling path scheme by using the following formula, based on the action output results and the adaptability evaluation results, combined with the resource capacity of the potential available resource set and the task priority order of the task priority sequence:

[0146]

[0147] in, For a given task Under the condition of allocating resources The probabilistic allocation quantity To the task priority The applied index is Nonlinear transformation, Indicates the task With resources Adaptability assessment results The applied index is a nonlinear transformation of for potentially available resources applying an exponential function a nonlinear transformation of for tasks a resource conflict probability between resources , , and are priority, fitness and potential available corresponding resource fitness adjustment factor respectively, is a policy temperature parameter, is a small constant to avoid division by zero, is an exponential function, is a set of all potentially available resources, k is the resource index, is a time-dependent decay function, where t is the current system time, is a resource suggested execution time window assigned to task .

[0148] In one embodiment, in the deduction simulation module 105, the preset burst scenario includes a resource overload scenario, a task emergency insertion scenario, and a resource temporary failure scenario;

[0149] Based on the preliminary resource allocation-scheduling path scheme, by simulating the resource supply and demand fluctuation and task priority mutation under the preset burst scenario, the adjustment process of the resource allocation and scheduling path is deduced, and the burst scenario simulation result is obtained, including:

[0150] For each preset burst scenario type, resource supply and demand fluctuation parameters and task priority mutation parameters under the corresponding scenario are generated, wherein the resource supply and demand fluctuation parameters include resource supply amount change parameters, resource demand amount change parameters under the corresponding scenario, and the task priority mutation parameters include new task priority parameters and original task priority adjustment parameters under the corresponding scenario.

[0151] Based on the preliminary resource allocation-scheduling path scheme, the resource supply and demand fluctuation data are matched with the resource type and the resource allocation amount corresponding to the resource type to determine the target resource;

[0152] Based on the preliminary resource allocation-scheduling path scheme, the task priority mutation parameters are matched with the node order information of the scheduling execution to determine the target task;

[0153] For each preset burst scenario type, based on the resource supply and demand fluctuation parameters, the allocation amount increase or decrease process of the target resource is deduced, and the target resource allocation amount adjustment scheme is obtained.

[0154] For each preset burst scenario type, the scheduling node insertion, deletion or reordering process of the target task is deduced based on the task priority mutation parameter to obtain a target task scheduling node change scheme.

[0155] The adjustment record of the target resource allocation amount, the target task scheduling node change record and the quantitative score comparison data before and after the task priority mutation for each preset burst scenario type are collected to form a burst scenario simulation result.

[0156] In one of the embodiments, the compensation optimization module 106 is further configured to:

[0157] Based on the burst scenario simulation result, the target task affected by the burst factor in the preliminary resource allocation-scheduling path scheme is identified.

[0158] According to the task priority sequence, the compensation resource allocation order of the target task is determined.

[0159] According to the compensation resource allocation order, the resource type allocated to the target task, the resource allocation amount of each resource type and the node order of scheduling execution are adjusted to form an optimized resource allocation-scheduling path scheme.

[0160] In one embodiment, a computer device is provided, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the real-time scheduling method of the dynamic priority-based integrated soc industrial computer as described above when executing the computer program.

[0161] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps in each method embodiment described above.

[0162] For the device embodiment, since it basically corresponds to the method embodiment, the related parts can be referred to the part of the method embodiment. The device embodiments described above are only schematic, and the components described as separate components can or can not be physically separate, and the components displayed as a unit can or can not be a physical unit, i.e. they can be located in one place or distributed on multiple network units. According to the actual needs, some or all of the modules can be selected to achieve the purpose of the present disclosure. Those skilled in the art can understand and implement it without creative labor.

[0163] The above-described embodiments only express several implementation manners of the application, the description is more specific and detailed, but it cannot be understood as the limitation of the patent scope of the application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the application, several modifications and improvements can be made, which are within the protection scope of the application.

Claims

1. A dynamic priority-based integrated soc industrial computer real-time scheduling method, characterized in that, The method comprises: Collecting system state data of the integrated soc industrial computer, the system state data comprising task state data and resource state data; Based on the resource state data, determining a set of potentially available resources by a state quantification method; Based on the task state data, evaluating the priority of each task in combination with a dynamic priority evaluation rule preset according to the business requirements of the industrial control scene, and generating a task priority sequence; Based on the set of potentially available resources and the task priority sequence, using a pre-trained reinforcement learning algorithm to generate a preliminary resource allocation and scheduling path scheme according to a preset state space and action space; Based on the preliminary resource allocation and scheduling path scheme, simulating the resource supply and demand fluctuations and task priority mutations under a preset burst scenario to deduce the adjustment process of the resource allocation and scheduling path, and obtaining a burst scenario simulation result; According to the burst scenario simulation result, using a priority-based compensation strategy to optimize the preliminary resource allocation and scheduling path scheme, and obtaining an optimized resource allocation and scheduling path scheme; According to the optimized resource allocation and scheduling path scheme, generating an integrated soc industrial computer scheduling instruction, which is used to drive the integrated soc industrial computer to execute tasks according to the optimized resource allocation logic and scheduling path.

2. The method of claim 1, wherein, Based on the resource state data, the set of potentially available resources is determined by a state quantification method, comprising: Classifying the resource state data by resource type to obtain a resource classification result, the resource type including computing resources, storage resources, and communication resources; For each type of resource in the resource classification result, configuring a state quantification index based on the hardware configuration parameters of the integrated soc industrial computer and the regular running load; Based on the state quantification index, quantifying the resource state data of each type of resource to obtain quantified values of each type of resource; Based on the quantified values of each type of resource, filtering out resources whose quantified values meet the corresponding preset available resource threshold to form the set of potentially available resources.

3. The method of claim 1, wherein, The dynamic priority evaluation rule includes the weight configuration of the key parameters of each task under different industrial control business scenarios; Based on the task state data, the priority of each task is evaluated in combination with a dynamic priority evaluation rule preset according to the business requirements of the industrial control scene, and a task priority sequence is generated, comprising: From the task state data, extracting key parameters associated with tasks under the industrial control scene, the key parameters including task deadline, data processing level, associated device dependency, and task execution urgency identifier; Based on the dynamic priority evaluation rule, the key parameters of each task are weighted calculated with the corresponding weight to obtain a priority quantification score of each task; The priority quantification scores of all tasks are sorted in descending order, and the task priority sequence is generated according to the sorting result.

4. The method of claim 1, wherein, In the pre-trained reinforcement learning algorithm, the state space is composed of the feature representation of the set of potentially available resources and the feature representation of the task priority sequence, and the action space is a set of feasible resource allocation methods and scheduling path types; The pre-training reinforcement learning algorithm is used to generate a preliminary resource allocation-scheduling path scheme according to a preset state space and an action space based on the set of potential available resources and the task priority sequence, including: The set of potential available resources and the task priority sequence are input into the state space for feature encoding to form initial state data; Based on the initial state data, a corresponding resource allocation mode and scheduling path type are matched in the action space to obtain an action output result; The initial state data and the action output result are adaptively evaluated by using the resource adaptation adjustment factor to generate an adaptability evaluation result; Based on the action output result and the adaptability evaluation result, the resource capacity of the set of potential available resources and the task priority order of the task priority sequence are combined to generate the preliminary resource allocation-scheduling path scheme, which includes the resource type allocated to each task, the resource allocation amount of each resource type, and node sequence information for scheduling execution.

5. The method of claim 4, wherein, Based on the action output result and the adaptability evaluation result, the resource capacity of the set of potential available resources and the task priority order of the task priority sequence are combined to generate the preliminary resource allocation-scheduling path scheme, which includes the resource type allocated to each task, the resource allocation amount of each resource type, and node sequence information for scheduling execution. wherein, for a given task under given conditions, the probabilistic allocation amount of resources , for the priority of the task , applies an exponential non-linear transformation with index , represents the adaptability evaluation result of the task to the resources , applies an exponential non-linear transformation with index , for the potential available resources applies an exponential non-linear transformation with index , is the resource conflict probability between the task and the resources , , and are the priority, adaptability and potential available corresponding resource adaptation adjustment factors respectively, is the strategy temperature parameter, is a small constant to avoid division by zero, is the exponential function, is the set of all potential available resources, k is the resource index, is a time-dependent decay function, where t is the current system time, is the suggested execution time window of the resources to the task .

6. The method of claim 4, wherein, The preset burst scenarios include resource overload scenarios, task emergency insertion scenarios, and resource temporary failure scenarios; Based on the preliminary resource allocation-scheduling path scheme, resource supply and demand fluctuations and task priority mutations in preset burst scenarios are simulated to deduce the adjustment process of resource allocation and scheduling path, and burst scenario simulation results are obtained, including: For each of the preset burst scenario types, resource supply and demand fluctuation parameters and task priority mutation parameters in the corresponding scenario are generated, wherein the resource supply and demand fluctuation parameters include resource supply amount change parameters and resource demand amount change parameters in the corresponding scenario, and the task priority mutation parameters include new task priority parameters and original task priority adjustment parameters in the corresponding scenario; Based on the preliminary resource allocation-scheduling path scheme, the resource supply and demand fluctuation data are matched with the resource types and the resource allocation amounts corresponding to the resource types to determine target resources; Based on the preliminary resource allocation-scheduling path scheme, the task priority mutation parameters are matched with the node sequence information for scheduling execution to determine target tasks; For each of the preset burst scenario types, the allocation amount increase or decrease process of the target resources is deduced based on the resource supply and demand fluctuation parameters to obtain a target resource allocation amount adjustment scheme; For each of the preset burst scenario types, the scheduling node insertion, deletion, or ordering re-adjustment process of the target tasks is deduced based on the task priority mutation parameters to obtain a target task scheduling node change scheme; The adjustment records of the target resource allocation amount, the target task scheduling node change records, and the quantitative score comparison data before and after the task priority mutation for each of the preset burst scenario types are collected to form the burst scenario simulation results.

7. The method of claim 6, wherein, The compensation optimization module is configured to, according to the burst scenario simulation result, adopt a priority-based compensation strategy to optimize the preliminary resource allocation-scheduling path scheme, and obtain an optimized resource allocation-scheduling path scheme. The compensation optimization module is configured to, according to the burst scenario simulation result, adopt a priority-based compensation strategy to optimize the preliminary resource allocation-scheduling path scheme, and obtain an optimized resource allocation-scheduling path scheme. The compensation optimization module is configured to, according to the burst scenario simulation result, adopt a priority-based compensation strategy to optimize the preliminary resource allocation-scheduling path scheme, and obtain an optimized resource allocation-scheduling path scheme. The system comprises:

8. A dynamic priority-based integrated soc industrial computer real-time scheduling system, characterized in that, The multi-source state acquisition module is configured to acquire system state data of the integrated SOC industrial computer, wherein the system state data comprises task state data and resource state data. The available resource evaluation module is configured to determine a set of potential available resources based on the resource state data by using a state quantification method. The priority evaluation module is configured to evaluate the priority of each task based on the task state data and in combination with a dynamic priority evaluation rule preset according to the business requirements of the industrial control scene, and generate a task priority sequence. The scheduling scheme generation module is configured to generate a preliminary resource allocation-scheduling path scheme according to a preset state space and action space by using a pre-trained reinforcement learning algorithm based on the set of potential available resources and the task priority sequence. The compensation optimization module is configured to, according to the burst scenario simulation result, adopt a priority-based compensation strategy to optimize the preliminary resource allocation-scheduling path scheme, and obtain an optimized resource allocation-scheduling path scheme. The scheduling instruction generation module is configured to generate an integrated SOC industrial computer scheduling instruction according to the optimized resource allocation-scheduling path scheme, wherein the integrated SOC industrial computer scheduling instruction is used to drive the integrated SOC industrial computer to execute tasks according to the optimized resource allocation logic and scheduling path. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 7. The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 7. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. ​ 10. A computer-readable storage medium having stored thereon a computer program, characterized in that, ​

Images