Semi-physical simulation resource scheduling method and device and storage medium

Abstract

The invention relates to the field of semi-physical simulation, in particular to a semi-physical simulation resource scheduling method and device and a storage medium. The method comprises the steps of obtaining a plurality of simulation models, a plurality of simulation devices and a plurality of test tasks in a semi-physical simulation resource library; combining the simulation models with different simulation devices to obtain semi-physical simulation nodes, and generating a plurality of cooperative combinations to cooperatively execute a plurality of test tasks to obtain performance test data of each simulation model; analyzing performance test data of each simulation model, and generating a performance parameter set of each simulation model, the performance parameter set comprising a plurality of scene sensitive indexes and a plurality of static basic indexes; and selecting a target simulation model and target simulation equipment from the semi-physical simulation resource library based on the simulation model requirements and the performance parameter set of each simulation model so as to execute a target simulation task. The scheduling strategy is highly adaptive to the actual operation environment, and the resource utilization rate and the scheduling accuracy are improved.

Description

Technical Field

[0001] The present application relates to the field of semi-physical simulation, and in particular to a scheduling method, device and storage medium for semi-physical simulation resources. Background Art

[0002] Hardware-in-the-Loop (HIL) technology provides an efficient and reliable means for the development, testing and verification of complex systems by organically combining physical devices with digital simulation models. Its application scenarios cover aerospace, automotive electronics, power systems, industrial automation and other fields, which can significantly reduce R&D costs and shorten product launch cycles. If advanced resource scheduling technology can be combined with the HIL framework, the efficiency and effectiveness of simulation can be further improved.

[0003] For example, the invention patent application with patent application publication number CN116774978A discloses a service-oriented reusable framework and integration method for simulation models. The service-oriented reusable framework for simulation models includes a resource layer, a service component layer, a service layer and an application layer; the resource layer is used to solve functions such as resource description, organization and management; the service component layer is used to encapsulate models of different levels, different fields, different categories and different granularities in the form of components; the service component layer mainly includes component description, component definition, component building tools, component verification, component configuration, component device, component reconstruction and component encapsulation, etc., and its main task is to complete the description, reconstruction and assembly of the model; the service layer is used to implement the encapsulation of the model in the form of cloud services, so as to shield the heterogeneity between simulation service models of different levels, different granularities and different fields. For another example, an invention patent application with patent application publication number CN119088562A discloses a simulation resource pool management and scheduling method and system, and a simulation method and system, wherein the simulation resource pool management and scheduling method includes: constructing a device resource pool and a model resource pool, wherein the device resource pool is configured with relevant information of multiple simulation devices, and the model resource pool is configured with runtime time information of multiple models; establishing a simulation task, and obtaining the runtime time information of the model corresponding to the simulation task from the model resource pool; calculating the priority value of the matching between the simulation device and the model in the device resource pool, calling the simulation device based on the optimal priority value, and allocating the called simulation device to the corresponding simulation task.

[0004] However, in order to achieve standardized management, existing resource scheduling technologies often use unified data encapsulation, which is difficult to meet the joint scheduling requirements of simulation models and physical devices in semi-physical simulation, resulting in insufficient dynamic resource allocation and flexible scheduling capabilities, limiting the efficiency and scalability of the simulation system. Summary of the invention

[0005] The main purpose of this application is to provide a scheduling method, device and storage medium for semi-physical simulation resources. In order to solve the above-mentioned technical problems, this application specifically adopts the following technical solutions: A first aspect of the present application is to provide a scheduling method for semi-physical simulation resources, the method comprising: S101, obtaining a semi-physical simulation resource library and test tasks under multiple different task scenarios, wherein the semi-physical simulation resource library includes a plurality of simulation models and a plurality of simulation devices; S102, combining each simulation model with different simulation devices to obtain a plurality of semi-physical simulation nodes, wherein the simulation models or simulation devices of the plurality of semi-physical simulation nodes are different; S103, generating a plurality of collaborative combinations based on the plurality of semi-physical simulation nodes to collaboratively execute a plurality of test tasks, and obtaining performance test data of each simulation model under different simulation devices, different collaborative combinations, and different task scenarios; S104, analyzing the performance test data of each simulation model to generate a performance parameter set for each simulation model, wherein the performance parameter set includes a plurality of scene-sensitive indicators and a plurality of static basic indicators; S105. Based on the task scenario and simulation model requirements of the target simulation task, and according to the simulation model requirements and the performance parameter set of each simulation model, a target simulation model and a target simulation device are selected from the semi-physical simulation resource library to execute the target simulation task.

[0006] In some embodiments, the performance test data includes several evaluation indicators; S104 includes: comparing multiple values ​​of the same evaluation indicator in the performance test data of each simulation model to determine the value fluctuation; when the value fluctuation is greater than a preset fluctuation value, determining the corresponding evaluation indicator as the scene-sensitive indicator, and storing different values ​​of the scene-sensitive indicator in association with the task scene and / or simulation equipment and / or collaborative combination; when the value fluctuation is less than the preset fluctuation value, determining the corresponding evaluation indicator as the static basic indicator.

[0007] In some embodiments, the method also includes: analyzing the changing trends of several values ​​of scene-sensitive indicators, and respectively determining the influence coefficients of task scenarios, simulation equipment, and collaborative combinations on the scene-sensitive indicators; determining the influence factors of the scene-sensitive indicators from the task scenarios, simulation equipment, and collaborative combinations according to the influence coefficients, and storing the influence factors in association with the scene-sensitive indicators.

[0008] In some embodiments, the method further includes: the static basic indicator is stored as a fixed value or a range interval, wherein the range interval is determined based on multiple values ​​of the static basic indicator.

[0009] In some embodiments, the simulation model requirements include performance requirements, and 105 also includes: based on the performance requirements and the static basic indicators of the simulation model, screening several unused simulation models to obtain several alternative simulation models; generating multiple alternative simulation nodes based on several of the alternative simulation models and several unused alternative simulation devices in the semi-physical simulation resource library; and obtaining multiple alternative collaboration combinations based on multiple combinations of the alternative simulation nodes; based on the scenario-sensitive indicators of the alternative simulation models, evaluating the predicted performance and resource occupancy rate of each of the alternative collaboration combinations in the task scenario of the target simulation task; selecting the alternative collaboration combination whose predicted performance meets the performance requirements and whose resource occupancy rate is the lowest as the target collaboration combination, and using the alternative simulation models and alternative simulation devices in the target collaboration combination as the target simulation model and the target simulation device.

[0010] In some embodiments, before S105, the method further includes: obtaining the priorities of the simulation tasks to be executed, and taking at least one simulation task with the highest priority as the target simulation task.

[0011] In some embodiments, the method further includes: loading the target simulation model into a corresponding target simulation device to obtain a number of target semi-physical simulation nodes; in the process of executing the target simulation task, continuously monitoring the task execution quality of a number of target semi-physical simulation nodes; when the task execution quality is lower than a preset quality, obtaining the measured performance data of the target semi-physical simulation node; if the difference between the measured performance data of the target semi-physical simulation node and the performance test data is greater than a preset difference, isolating the simulation data generated by the target semi-physical simulation node.

[0012] In some embodiments, the method also includes: if the difference between the measured performance data and the performance test data of the target semi-physical simulation node is less than the preset difference, based on the scenario-sensitive indicators of the target simulation model, selecting a preferred simulation device from several unused simulation devices in the semi-physical simulation resource library; loading the target simulation model into the preferred simulation device to replace the target simulation device, and updating the target semi-physical simulation node.

[0013] A second aspect of the present application is to provide a computer device, the device comprising: Memory for storing computer programs; A processor is used to execute the computer program and implement the steps of the scheduling method of semi-physical simulation resources provided in any embodiment of the present application when executing the computer program.

[0014] The third aspect of the present application is to provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor performs the steps of the scheduling method for semi-physical simulation resources provided in any embodiment of the present application.

[0015] Beneficial effects: The embodiment of the present application provides a scheduling method, device and storage medium for semi-physical simulation resources. Through a diverse combination of heterogeneous resource tests, the performance parameters of each simulation model are centralized to establish a dynamic and static separation index system, and the model performance is divided into scenario-sensitive indicators and static basic indicators and customized storage to show the performance boundary and dynamic impact relationship of the simulation model in different environments. Thus, the strong coupling between semi-physical simulation resources is transformed into an advantage in improving collaborative efficiency, so that it can achieve a high degree of adaptation between the scheduling strategy and the actual operating environment based on the multi-dimensional analysis of the performance parameter set, thereby improving resource utilization and scheduling accuracy.

[0016] Specifically, by cross-combining simulation models with different simulation devices to generate diversified simulation nodes and executing test tasks in multi-task scenarios, complex scenarios under different test conditions (such as task scenario type, simulation device model, collaborative combination configuration) are covered, thereby capturing the synergy and implicit dependencies between digital models and physical hardware, and different model combinations. Furthermore, by comparing performance parameter sets, the significantly affected indicators are identified, dynamic and static indicators are separated, and dynamic indicators are associated with their key influencing factors for storage, breaking the "one-size-fits-all" data encapsulation method and providing a basis for dynamic scheduling.

[0017] Furthermore, in the early stage of task matching, models that do not meet the basic capability requirements are quickly filtered out through static indicators to reduce the computational complexity of subsequent dynamic evaluation. Then, scenario-sensitive indicators are used to quantify the real-time impact of task scenarios, equipment characteristics, and collaboration modes on model performance. Based on multiple dimensions (simulation performance, resource occupancy), layer-by-layer screening and overall screening mechanisms are used to implement resource scheduling, respond to performance degradation or complementary effects in a strongly coupled environment, predict and avoid performance traps, break through the limitations of single resource performance evaluation, and optimize the dynamic response capabilities of resource scheduling to complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale. Obviously, the drawings described below are some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without paying creative labor.

[0019] Figure 1 is a schematic flow chart of a scheduling method for semi-physical simulation resources provided in an embodiment of the present application; Figure 2 It is a schematic diagram of a semi-physical simulation resource library package provided in an embodiment of the present application; Figure 3 It is a schematic block diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0021] Herein, suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present application, and have no specific meanings by themselves. Therefore, "module", "component" or "unit" can be used mixedly.

[0022] In this document, the terms "upper", "lower", "inner", "outer", "front", "back", "one end", "the other end" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0023] Herein "and / or" includes any and all combinations of one or more of the associated listed items.

[0024] Herein, "plurality" means two or more than two, ie, it includes two, three, four, five, etc.

[0025] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0026] Semi-physical simulation is a technology that combines actual physical hardware with computer-generated digital models for system testing. It verifies the performance and reliability of the system by running real hardware and virtual models in a closed-loop environment.

[0027] In this article, simulation models refer to mathematical or logical models used to simulate system behavior, that is, virtual models in semi-physical simulation. These models can be dynamic models based on physical laws, control system algorithms, or signal processing processes. They are usually implemented in simulation software and can respond to input data and generate output results.

[0028] In this article, simulation devices refer to actual physical components that directly participate in the experimental process, such as sensors, actuators, controllers, etc., that is, real hardware devices in semi-physical simulation. These hardware devices receive data from the simulation model as input and convert it into real physical actions or state changes; at the same time, they can also feedback the results of actual operations to the simulation model.

[0029] In a semi-physical simulation system, there is a close connection between the simulation model and the simulation device. On the one hand, the simulation model provides control instructions or environmental parameters for the physical device; on the other hand, the physical device performs corresponding operations based on the received information and feeds back the results to the simulation model to update its status. This interaction requires a high degree of consistency and synchronization between the two. The strong coupling between the simulation model and the physical device also leads to mutual influence between the two. In addition, with the expansion of system scale and diversification of application scenarios, simulation models in different fields often need to work together to form a joint simulation of heterogeneous models, which will also affect each other. For example, the drone model completes the observation task and the radar model completes the detection task. The two simulation models work together to complete the task, and the performance of the two simulation models has a coupling relationship.

[0030] Existing resource scheduling technologies tend to adopt a unified data encapsulation method and standardize data formats for easy storage and retrieval, but they have serious limitations when facing cross-device and cross-domain scenarios. Due to the lack of a flexible adaptation mechanism, different types of data are difficult to seamlessly connect, which easily forms information islands. Even with a rich resource library, it is impossible to accurately implement dynamic resource allocation and flexible scheduling, which limits the efficiency and scalability of the simulation system.

[0031] Based on this, the embodiments of the present application provide a scheduling method, device and storage medium for semi-physical simulation resources. Through a diversified combination of heterogeneous resource tests, the performance parameters of each simulation model are centralized to establish a static and dynamic separation index system, and the model performance is divided into scenario-sensitive indicators and static basic indicators and customized storage to show the performance boundary and dynamic impact relationship of the simulation model in different environments. Among them, the static basic indicators quantify the inherent performance of the model and are used to screen simulation models with basic capabilities that meet the standards. The scenario-sensitive indicators dynamically reflect the performance fluctuations and adaptability of the model in the real deployment environment, and are used for joint simulation performance evaluation. In this way, the strong coupling between semi-physical simulation resources is transformed into an advantage in improving collaborative efficiency, so that it can flexibly match task requirements and resource characteristics based on multi-dimensional analysis of the performance parameter set, and predict and avoid performance traps.

[0032] Some embodiments of the present application are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features of the embodiments can be combined with each other. Figure 1 , Figure 1 is a schematic flow chart of a scheduling method for semi-physical simulation resources provided in an embodiment of the present application, such as Figure 1 As shown, an embodiment of the present application provides a scheduling method for semi-physical simulation resources, and the method includes S101 to S105.

[0033] S101, obtaining a semi-physical simulation resource library and test tasks in multiple different task scenarios, wherein the semi-physical simulation resource library includes a plurality of simulation models and a plurality of simulation devices.

[0034] Among them, the semi-physical simulation resource library refers to a comprehensive resource pool that stores a variety of simulation models (such as aircraft models, radar models, UAV models, satellite models, etc.) and a variety of simulation equipment (such as real-time simulators, five-axis flight turntables, torque load simulators, flight control computers, external mission computers, GPS satellite navigation generators, etc.).

[0035] Specifically, multi-dimensional mission scenarios are defined to cover a combination of different mission types (such as detection missions, collaborative missions) and different mission background environments (such as mountainous terrain, severe weather conditions, dense urban buildings, and electromagnetic interference). Each scenario corresponds to specific simulation conditions and constraints. Correspondingly, the test tasks are specific verification requirements designed for different mission scenarios, such as testing the communication stability of a UAV formation in a rainy environment, or evaluating radar detection accuracy in an urban canyon scenario. For example, a limited number of representative mission scenarios can be pre-set according to actual needs. On the one hand, it comprehensively covers the collaborative adaptability of models and equipment under diverse conditions. On the other hand, it avoids redundant or invalid test tasks that increase the burden of early testing, and provides accurate and reliable basic data for subsequent performance analysis. Specific mission scenarios and test tasks are not limited here.

[0036] It should be understood that by simulating the complexity of real scenarios, the adaptability and collaborative effectiveness of simulation resources in heterogeneous environments can be systematically verified. For example, detection tasks can test the attenuation of the model's perception accuracy in extreme weather, and collaborative interaction tasks can test the fluctuation of multi-device communication delays under electromagnetic interference. This multi-dimensional test can obtain fine-grained dynamic performance data, provide a differentiated scenario adaptation basis for subsequent resource scheduling, and avoid performance evaluation deviations caused by single scenario testing.

[0037] S102: Combine each simulation model with different simulation devices to obtain a plurality of semi-physical simulation nodes, wherein the simulation models or simulation devices of the plurality of semi-physical simulation nodes are different.

[0038] Specifically, a semi-physical simulation node refers to the smallest execution unit composed of a simulation model and at least one simulation device, where the simulation model is responsible for simulating the digital logic of the system under test, and the simulation device provides a physical interface or hardware environment. The two work together to complete a specific simulation task. By fully cross-combining the simulation model and the simulation device, a variety of semi-physical simulation nodes are generated. For example, aircraft model A can be combined with a flight control device and a five-axis flight turntable to obtain two semi-physical simulation nodes.

[0039] It should be understood that a variety of semi-physical simulation nodes cover the diverse possibilities of models and devices in hardware and software collaboration, forming a complete set of "model-device" pairings. Through the differences between nodes, the adaptability of the model on different devices is tested, providing a basis for subsequent analysis of the effectiveness of software and hardware collaboration. Exemplarily, after the simulation model is loaded into the simulation device according to the "model-device" pairing in the semi-physical simulation node, it can be used to execute test tasks or simulation tasks.

[0040] In some embodiments, the simulation device can be a single hardware device or a hardware system integrated by multiple hardware devices. That is to say, multiple hardware devices can be integrated in each simulation device, and the simulation device is divided into lightweight nodes and fully quantized nodes according to the depth and breadth that the simulation device can simulate. Among them, lightweight nodes can be used to simulate or replace the functions of certain small controllers or sensors in the system, and fully quantized nodes are closer to complete hardware systems in actual applications, providing more comprehensive and accurate functional simulations than lightweight nodes or directly running as part of a more reliable system.

[0041] For example, in the simulation of unmanned aerial vehicles, lightweight nodes can include real-time simulators, flight control computers, and external mission computers, which are used for rapid verification of core functions to meet the requirements of lightweight and low-cost simulation; fully quantitative nodes can include real-time simulators, five-axis flight turntables, torque load simulators, flight control computers, external mission computers, GPS satellite navigation generators, etc. The configuration of fully quantitative nodes is more comprehensive and can more accurately simulate the execution of tasks such as flight and detection in real environments. Among them, the real-time simulator is the core computing unit, running the corresponding simulation model, environmental model and control algorithm; the flight control computer is used to execute flight control logic (such as attitude control, navigation algorithm); the external mission computer is used to manage the simulation of mission loads (such as sensors, weapon systems); the five-axis flight turntable is used to simulate the attitude changes of the aircraft (such as pitch, roll, yaw, X / Y / Z axis displacement); the torque load simulator is used to simulate the dynamic response of the aircraft under aerodynamic and mechanical loads.

[0042] It should be understood that semi-physical simulation will also use various integrated hardware systems corresponding to lightweight nodes and fully quantitative nodes in actual applications. The device combination rules can be set in advance based on actual usage requirements. If the simulation device is a single hardware device, multiple hardware devices can be combined into simulation devices in the form of multiple hardware systems according to the device combination rules.

[0043] As a result, the combination of each simulation model and simulation equipment is both diversified and streamlined, and ultimately a controllable number of semi-physical simulation nodes are formed. On the one hand, it is more in line with the actual configuration of the semi-physical simulation nodes in practical applications, which improves the effectiveness of the semi-physical simulation nodes used in the test, and thus improves the availability of subsequent performance test data; on the other hand, there is no need to perform tedious and complex high-precision pairing between the simulation model and the simulation equipment according to other factors such as the performance and task scenarios required to be tested by the simulation model, thereby avoiding redundant or invalid semi-physical simulation nodes from generating interfering test data, and effectively reducing the complexity of semi-physical simulation node pairing and the difficulty and computational complexity of subsequent testing.

[0044] S103, generating a plurality of collaborative combinations based on the plurality of semi-physical simulation nodes to collaboratively execute a plurality of test tasks, and obtaining performance test data of each simulation model under different simulation devices, different collaborative combinations, and different task scenarios.

[0045] Specifically, a collaborative combination refers to a collaborative working unit composed of two or more semi-physical simulation nodes, which is used to simulate the scenario of multiple devices jointly executing tasks during a real task. For each test task, several semi-physical simulation nodes can be selected from a variety of semi-physical simulation nodes to generate multiple collaborative combinations according to the simulation model requirements of the test task. The corresponding semi-physical simulation nodes are called multiple times through different collaborative combinations to verify the performance of each simulation model under different simulation devices, different collaborative combinations, and different task scenarios, and the performance test data of each simulation model is recorded. In other words, the performance test data is the model performance data of the simulation model under specific conditions, such as the detection accuracy of the radar model B loaded into the full quantization node and the lightweight node respectively in the task scenario of electromagnetic interference or sunny scene, and the calculation time of a single reasoning.

[0046] For example, in UAV simulation, the collaborative combination can be a cluster or a team in the cluster responsible for performing a specific task. The collaborative combination can include the first semi-physical simulation node "aircraft model A + full-quantized node", the second semi-physical simulation node "radar model B + lightweight node", and the third semi-physical simulation node "satellite model C + lightweight node".

[0047] For example, when the test task is a detection task, the detection task needs to use the search, tracking, time synchronization and other capabilities of the drone, and the simulation model requirement of the test task is 3 drone models. Multiple semi-physical simulation nodes corresponding to the drone models are screened out, and based on these semi-physical simulation nodes, multiple collaborative combinations of 3 semi-physical simulation nodes as a group are generated to execute the test task separately.

[0048] In some embodiments, different or identical semi-physical simulation nodes can be combined into a collaborative combination to verify the overall performance of homogeneous or heterogeneous resources working together. That is, several semi-physical simulation nodes in the collaborative combination can be completely identical, partially identical, or completely different, which is not limited here.

[0049] In some embodiments, whether it is a test task or a target simulation task to be performed, it includes several elements such as the use requirements of the simulation objects, task objectives, environmental parameters, time constraints, and interaction rules. Among them, the task type can be determined according to the task objectives, and the general task scene can be predicted in combination with some environmental parameters as public information, such as the weather conditions and terrain conditions at the task location. Furthermore, the simulation model requirements can be comprehensively analyzed based on the multiple elements of the task, and the simulation model requirements can include the quantity requirements, type requirements, and performance requirements of the simulation model for the task.

[0050] For example, if the use requirement of the simulation object is to collaboratively perform the tracking task of multiple target objects within three drones, it can be determined that the use requirement of the simulation model is 3 (i.e. quantity requirement) drone models (i.e. type requirement), and the communication delay index of the drone is required to be less than or equal to 50ms (i.e. performance requirement). For another example, if the use requirement of the simulation object is to collaboratively perform the tracking task of multiple target objects within a drone, it can be determined that the use requirement of the simulation model is a drone model (i.e. type requirement), and the communication delay index of the drone is required to be less than or equal to 50ms (i.e. performance requirement).

[0051] For example, based on the task type, task background, and usage requirements of the simulation object, key performance requirements are extracted. For example, detection tasks require high-precision perception, collaborative tasks require low-latency communication, and mountain scenes require strong anti-interference capabilities. Then, specific thresholds are set in combination with task execution conditions (such as real-time requirements and resource budget constraints) to obtain comprehensive performance requirements. It should be noted that the task scenarios and simulation model requirements of both the test task and the target simulation task to be executed can be determined and modified by the user, or obtained by referring to relevant technologies, and are not limited here.

[0052] It should be understood that for test tasks, their environmental parameters can be stored as task scenarios in association with corresponding scenario-sensitive indicators, and their simulation model requirements only consider the quantity and type requirements of the task for the simulation model. For target simulation tasks, only some basic information of their environmental parameters is public information, and can be gradually disclosed as the simulation task progresses (such as environmental data obtained after executing the detection task). It is necessary to fully consider the quantity, type and performance requirements of the simulation model, and the performance requirements can be matched to the simulation model that meets the requirements based on the performance parameter set.

[0053] S104, analyzing the performance test data of each simulation model, and generating a performance parameter set for each simulation model, wherein the performance parameter set includes a plurality of scene-sensitive indicators and a plurality of static basic indicators.

[0054] Specifically, all collected performance test data are statistically analyzed, and the value changes of the same evaluation indicator under various conditions are compared to determine the stability of the performance indicator under different task scenarios, simulation equipment types or collaborative combinations. Different evaluation indicators are then stored based on stability differences, generating a performance parameter set that comprehensively reflects its performance characteristics, and systematically describing the performance of each simulation model under various operating conditions.

[0055] Among them, static basic indicators are the inherent and relatively stable stable model performance of the model, reflecting the basic performance of the single simulation model. For example, the technical characteristics of the simulation model, such as the operating frequency, signal processing capability, and data processing capability, remain relatively stable under different conditions. This type of indicator is less affected by the external environment and has less fluctuations under different simulation equipment or mission scenarios and collaborative combinations. For example, the "minimum detection distance" of a radar model has been stable at around 5 meters in multiple tests.

[0056] Among them, scene-sensitive indicators are dynamic model performance that changes significantly with external environment or collaborative conditions. For example, the specific implementation-level characteristics of the radar model, such as the range, coverage, detection accuracy, and bit error rate, may change significantly with changes in the environment (such as increased electromagnetic interference intensity or increased terrain complexity). Such indicators have dynamic performance fluctuations in specific mission scenarios, simulation devices, or collaborative combinations. For example, although simulation model D has high accuracy on simulation device X, its collaboration with simulation model E will cause a surge in communication delays due to resource competition; for example, the data packet loss rate of the communication model may surge from 0.1% in a normal environment to 5% in an electromagnetic interference scenario, or the path planning success rate of the navigation model may decrease in lightweight nodes in complex terrain due to insufficient computing resources.

[0057] In some embodiments, scenario-sensitive indicators include: collaboration indicators, which are used to characterize the overall effectiveness of multi-node collaborative work, such as data synchronization delay between nodes, task collaborative execution success rate, and resource competition conflicts when multiple models run concurrently; scenario indicators, which are used to characterize the impact of environmental dynamics on performance, such as terrain matching errors in mountainous scenarios, communication bit error rates in strong electromagnetic interference, and the number of re-planning times of path planning algorithms in multi-obstacle environments.

[0058] It should be understood that the performance parameter set provides the basic performance framework of the simulation model through static basic indicators, and also reflects how the simulation model responds to external factors in a complex and changeable actual operating environment through scene-sensitive indicators. The final performance parameter set decouples the static capability and dynamic adaptability of the simulation model, providing a multi-dimensional decision-making basis for subsequent scheduling, ensuring that resources and needs can be accurately matched during dynamic scheduling. For example, in the performance parameter set of the "visual recognition model", the static basic indicators include "image processing resolution 1080p", while the scene-sensitive indicators include "recognition accuracy under low light conditions drops from 92% to 78%, and can be increased to 85% when linked with infrared sensor device E", thereby providing a quantitative basis for subsequent resource scheduling that takes into account stability and flexibility.

[0059] In some embodiments, the performance parameter set of each simulation model is associated with a type label of the simulation model, such as a radar model, an aircraft model, a drone model, etc., which is used to quickly screen out the required type of simulation model and its performance parameter set according to the simulation model requirements.

[0060] In some embodiments, the performance test data includes several evaluation indicators. Among them, the evaluation indicators are dimensions used to quantify the performance of the simulation model under specific test conditions (such as task scenario type, simulation equipment model, collaborative combination configuration). The evaluation indicators are reflected by the specific numerical values ​​of the performance test data. Different types of simulation models can use different evaluation indicators. These evaluation indicators include static characteristics inherent to the model and dynamic characteristics significantly affected by the external environment.

[0061] In some embodiments, S104 includes: comparing multiple values ​​of the same evaluation indicator in the performance test data of each simulation model to determine the value fluctuation; when the value fluctuation is greater than the preset fluctuation value, determining the corresponding evaluation indicator as the scene-sensitive indicator, and storing different values ​​of the scene-sensitive indicator in association with the task scene and / or simulation equipment and / or collaborative combination; when the value fluctuation is less than the preset fluctuation value, determining the corresponding evaluation indicator as the static basic indicator.

[0062] Among them, value fluctuation refers to the range of value changes of the same evaluation indicator in different task scenarios, simulation equipment or collaborative combinations. The value fluctuation can be quantified by standard deviation, range or relative change rate, which is not limited here. Correspondingly, a preset fluctuation value can be set to determine the stability of the evaluation indicator, and its specific value can be set according to actual needs.

[0063] Specifically, the types of evaluation indicators are distinguished by quantitatively analyzing the range of value changes of the same evaluation indicator in the execution of multiple test tasks in the performance test data. If the value fluctuation of a certain evaluation indicator is greater than the preset fluctuation value, indicating that its performance is significantly affected by the external environment, it is determined to be a scene-sensitive indicator, and its different values ​​are associated with specific influencing factors and stored, such as a specific type of task scene, simulation equipment model, and collaborative combination configuration, to form a dynamic performance mapping relationship. For example, the "bit error rate of 3.5%" of simulation model D is associated with "simulation equipment X" and "mountain scene". If the value fluctuation of a certain evaluation indicator is less than the preset fluctuation value, it is determined to be a static basic indicator, and its mean or stable interval is taken as a fixed parameter.

[0064] For example, the calculation time of a simulation model on three simulation devices is 15ms, 18ms, and 16ms respectively, the value fluctuation is 3ms, and the preset upper limit of the fluctuation value is 20% of the median. That is, the fluctuation within the range of 3.2ms above and below the median 16ms is a static basic indicator, and the range exceeding 3.2ms is a scene-sensitive indicator. At this time, it can be determined that the calculation time of a simulation model is an evaluation indicator as a static basic indicator.

[0065] Exemplarily, the standard deviation of the performance test data across scenarios for each evaluation indicator is calculated. If the standard deviation exceeds the preset fluctuation value, the scenario-sensitive flag is triggered and the associated metadata (such as simulation device ID, collaboration combination number) is extracted to construct a performance parameter set that has both stability description and dynamic adaptation relationship.

[0066] In some embodiments, it is characterized in that the method also includes: analyzing the changing trends of several values ​​of scene-sensitive indicators, and respectively determining the influence coefficients of task scenarios, simulation equipment, and collaborative combinations on the scene-sensitive indicators; determining the influence factors of the scene-sensitive indicators from the task scenarios, simulation equipment, and collaborative combinations according to the influence coefficients, and storing the influence factors in association with the scene-sensitive indicators.

[0067] Specifically, the change trend refers to the numerical change law of the same indicator under different test conditions (such as task scenario type, simulation equipment model, and collaborative combination configuration). For example, the "data packet loss rate" of the communication model increases linearly with the increase of electromagnetic interference intensity. According to the change trend, the independent influence of the three factors of task scenario, simulation equipment, and collaborative combination on the indicator is quantified, that is, the influence coefficient. Correspondingly, a preset coefficient can be set to determine the standard for a factor to have a significant impact on the indicator. The specific value can be flexibly set according to actual needs. When the influence coefficient of a factor is greater than the preset coefficient, the factor is marked as an influence factor and stored in association with the scenario-sensitive indicator.

[0068] Taking the "detection distance" evaluation index of a certain radar model as an example, when using simulation device A, it is 30% shorter than simulation device B on average, so the impact coefficient of the simulation device type is 0.3; and the "mountainous terrain" mission scenario causes the same index to drop by 20%, so the impact coefficient of the mission scenario type is 0.2; and the same index only fluctuates up and down by 8% in different collaborative combinations, so the impact coefficient of the collaborative combination type is only 0.08. If the preset coefficient is 0.1, the "collaborative combination" is not included in the influencing factor, and the influencing factors of the scene-sensitive index "detection distance" are "simulation device" and "mission scenario". Correspondingly, the "detection distance" of a certain radar model is stored in association with "simulation device" and "mission scenario".

[0069] For example, the identification and storage of influencing factors can be realized through automated data analysis tools, or the influencing factors can be quantified by using correlation analysis algorithms such as regression analysis or sensitivity analysis. The influencing factors of the same indicator can also be identified by controlling variables, for example, in the same task scenario and the same simulation device, the impact of collaborative combinations can be analyzed. The specific algorithm is not limited here.

[0070] In some embodiments, it is characterized in that the method further includes: the static basic indicator is stored as a fixed value or a range interval, wherein the range interval is determined based on multiple values ​​of the static basic indicator.

[0071] Specifically, for static basic indicators, a differentiated strategy is adopted according to their data stability: if the indicator has very small numerical fluctuations and highly consistent performance in cross-scenario tests, it will be stored as a fixed value, such as "memory usage 20.3%"; if there are slight fluctuations within a controllable range, the extreme value range or confidence interval is calculated by statistically analyzing multiple values ​​in the performance test data. For example, if the calculation time of the path planning model fluctuates between 10ms and 12ms, it will be stored as a range interval based on the statistical distribution, such as "time interval [10ms, 12ms]"; and the communication time of the communication model in the three collaborative combinations is 15ms, 15.5ms, and 14.8ms, respectively, and its range interval is defined as [14.5ms, 15.6ms].

[0072] In this way, the query efficiency of basic performance is simplified, and the necessary details are retained through the upper and lower limits of the interval, providing flexible matching space for subsequent scheduling, while avoiding the problem of false screening caused by overly strict single fixed values.

[0073] It should be understood that the final parameter set decouples static capabilities from dynamic adaptability. For example, in the parameter set of "visual recognition model D", static basic indicators include "image processing resolution 1080p", while scene-sensitive indicators are recorded in combination with the test environment context as "recognition accuracy under low-light conditions drops from 92% to 78%, and can be increased to 85% when linked with infrared sensor device E", thereby providing a quantitative basis for subsequent resource scheduling.

[0074] In some embodiments, before S105, the method further includes: obtaining the priorities of the simulation tasks to be executed, and taking at least one simulation task with the highest priority as the target simulation task.

[0075] Specifically, the priority of all pending simulation tasks is obtained. The priority can be defined by the user or automatically generated according to the urgency of the task, resource requirements, and deadline. For example, the priority of the rescue task is higher than that of the patrol task. It is specifically expressed as a numerical value or level. Then, one or more tasks with the highest priority are selected as "target simulation tasks" according to the priority label to ensure that critical tasks with high timeliness or high value are given priority in resource matching in resource competition scenarios, avoiding delays in core processes due to confusion in task order.

[0076] S105. Based on the task scenario and simulation model requirements of the target simulation task, and according to the simulation model requirements and the performance parameter set of each simulation model, a target simulation model and a target simulation device are selected from the semi-physical simulation resource library to execute the target simulation task.

[0077] Specifically, for the target simulation task to be performed, the task type can be determined according to the task goal, combined with the task scenario predicted by the currently disclosed environmental parameters, such as the weather conditions and terrain conditions at the task location. Furthermore, the simulation model requirements can be determined according to the use requirements of the simulation object, such as the performance requirements and type requirements of the simulation model for the target simulation task. First, the corresponding type of simulation model is screened out based on the type requirements, and then the known task scenario of the target simulation task is used as the limiting item, and the collaborative combination and simulation equipment are optional. The simulation model whose evaluation index meets the performance requirements is selected. If the simulation model requirements have a clear quantity requirement for a certain type of simulation model, finally, the corresponding number of target simulation models and target simulation equipment with matching relationships can be determined according to the quantity requirement of the simulation model for the target simulation task.

[0078] In some embodiments, the performance parameter set of each simulation model is screened layer by layer according to the simulation model requirements. First, the simulation models that do not meet the performance requirements are quickly filtered through static basic indicators; for qualified simulation models that meet the static basic indicators, combined with the task scenario of the target simulation task, the task scenario determined in the target simulation task is limited, and according to the task scenario and simulation equipment associated with the scene-sensitive indicators of the qualified simulation model, the specific values ​​of the scene-sensitive indicators that the simulation model may present under different task scenarios and simulation equipment are flexibly analyzed, and the simulation model and simulation equipment combination whose lower limit value meets the performance requirements is selected, that is, the qualified simulation node, to achieve strong binding screening between the simulation model and the simulation equipment, to ensure that the two can meet the static performance benchmark and adapt to the dynamic environment requirements when working together.

[0079] Therefore, the simulation model and the simulation device are regarded as a whole execution unit for joint evaluation, which fundamentally solves the problem of substandard performance caused by the decoupling of hardware and algorithms in traditional scheduling, and effectively avoids performance degradation combinations. For example, although a simulation model meets the static indicators, it will cause real-time crashes if combined with low-performance devices. Binding filtering can directly exclude such combinations, avoid the subsequent generation of invalid collaborative combinations, and reduce the computational workload and efficiency of resource scheduling.

[0080] Furthermore, based on screening out several qualified simulation nodes that meet the performance requirements, several qualified collaborative combinations are generated, and based on the collaborative combinations associated with the scenario-sensitive indicators of the simulation models, the synergy between several simulation models is evaluated. For example, some combinations will have better or worse performance when working together. At this time, the simulation models and simulation devices in the qualified collaborative combinations with the best synergy can be preferentially selected as the target simulation models and target simulation devices.

[0081] Therefore, firstly, simulation models whose static basic indicators do not meet the standards are excluded, and then the simulation models and simulation devices are screened as bound execution units for the second time to exclude simulation nodes whose scene-sensitive indicators do not meet the standards, and finally, collaborative combinations are introduced to screen for the optimal collaborative efficiency. Through layer-by-layer screening, the redundant computing amount brought by unqualified and inefficient simulation nodes and collaborative combinations is effectively reduced, thereby improving the efficiency and response speed of resource scheduling. It should be noted that for different collaborative combinations, the performance of the simulation model may fluctuate slightly downward due to resource competition and other reasons, but in actual applications, more cases are that mutual collaboration produces better model performance. Therefore, even if the screening of the collaborative combination is postponed, the selection of downward fluctuating collaborative combinations can be avoided based on the principle of selecting the best, ensuring that the target simulation model and target simulation device finally selected can still meet the performance requirements.

[0082] It should be understood that the performance parameters of each simulation model have established a static and dynamic separation indicator system, breaking the "one-size-fits-all" resource description model. Static basic indicators are independent of environmental changes and are used to reflect the inherent performance of the model. In the early stage of task matching, static indicators are used to quickly filter out models that do not meet the basic capability requirements, greatly reducing the computational complexity of subsequent dynamic evaluations. The scene-sensitive indicators quantify the performance boundaries of the model in real deployment by associating influencing factors such as task scenarios and equipment models. In the resource binding stage, the model-equipment combination with the best scene-sensitive indicators is given priority in combination with the current task scenario. In this way, the efficiency advantage of traditional resource scheduling is retained, and the dynamic response capability to complex environments is improved.

[0083] In some embodiments, since the task scenario can have various parameters, and the target simulation task can generally disclose or predict some task scenarios, but not all task scenarios, the scenario-limited parameters (such as weather conditions and temperature conditions) can be determined for known task scenarios, and the scenario-optional parameters can be determined for unknown task scenarios. In the subsequent model selection process, the scene-sensitive indicators can be pre-screened according to the parameter values ​​of the scene-limited parameters, and the scene-limited parameters stored in association with the scene-sensitive indicators match the scene-limited parameters of the target simulation task (for example, the weather condition of the target simulation task is cloudy, and the weather condition bound to a certain scene-sensitive indicator is also cloudy). The scene-limited parameters stored in association with the scene-sensitive indicators do not match the scene-limited parameters of the target simulation task (for example, the limiting parameters of the target simulation task are 20°C, and the weather condition bound to a certain scene-sensitive indicator is 30°C-40°C) are set to a hidden state. Thus, the known task scenario of the target simulation task is used as a limiting item to ensure that the performance parameters that do not match the task scenario of the target simulation task are not used in the resource allocation process of this target simulation task, thereby further improving the efficiency of resource allocation.

[0084] In some embodiments, the simulation model requirements include performance requirements, and 105 also includes: based on the performance requirements and the static basic indicators of the simulation model, screening several unused simulation models to obtain several alternative simulation models; generating multiple alternative simulation nodes based on several of the alternative simulation models and several unused alternative simulation devices in the semi-physical simulation resource library; and obtaining multiple alternative collaboration combinations based on multiple combinations of the alternative simulation nodes; based on the scenario-sensitive indicators of the alternative simulation models, evaluating the predicted performance and resource occupancy rate of each of the alternative collaboration combinations in the task scenario of the target simulation task; selecting the alternative collaboration combination whose predicted performance meets the performance requirements and whose resource occupancy rate is the lowest as the target collaboration combination, and using the alternative simulation models and alternative simulation devices in the target collaboration combination as the target simulation model and the target simulation device.

[0085] Specifically, the performance parameter set of each simulation model is screened in multiple dimensions according to the simulation model requirements. First, several unused simulation models in the semi-physical simulation resource library are quickly filtered through static basic indicators to remove simulation models that do not meet the standards. For several alternative simulation models that meet the static basic indicators, multiple alternative simulation nodes are generated based on several alternative simulation models and several alternative simulation devices that are not used in the semi-physical simulation resource library. On this basis, multiple alternative simulation nodes are randomly combined to obtain multiple alternative collaboration combinations. According to the collaboration combinations, task scenarios, and simulation devices associated with the scene-sensitive indicators of the alternative simulation models, combined with the task scenarios of the target simulation tasks, the specific values ​​of the scene-sensitive indicators that the simulation models may present under different alternative collaboration combinations and alternative simulation devices are flexibly analyzed, and then the comprehensive performance that can be generated by the collaboration of multiple simulation models is evaluated from the macro perspective of the entire collaboration combination, that is, the predicted performance. Among them, the predicted performance refers to the comprehensive performance level expected to be achieved by the alternative collaboration combination in the target simulation task scenario, such as task completion time, target recognition accuracy, and collaborative delay.

[0086] At the same time, resource occupancy can be comprehensively evaluated from the dimensions of the number of simulation models occupied by the alternative collaboration combination, the number of simulation devices, and the hardware resource consumption required to run the simulation (such as CPU occupancy, memory bandwidth, and communication bandwidth). The predicted performance and resource occupancy can then be converted into a comprehensive score through a weighted algorithm, such as giving priority to the target collaboration combination of "performance meets the standard and hardware resource consumption is the lowest" or "performance meets the standard and the number of simulation models is the least". In this way, on the one hand, the actual occupancy can be accurately calculated to avoid resource overload, and on the other hand, the simulation task can be selectively completed with fewer simulation resources. The saved simulation resources can be used for other simulation tasks to maximize resource utilization efficiency.

[0087] For example, for the alternative collaborative combination including "simulation model F+simulation device X" and "simulation model G+simulation device Y", by retrieving the scene sensitivity index of simulation model F as "85% recognition rate of strong electromagnetic interference scene on simulation device X" and the scene sensitivity index of simulation model G as "mountain path planning on simulation device Y takes 18ms", combined with the static basic index of simulation model F and simulation model G as "data synchronization interval (5ms)", the prediction model is used to predict that the comprehensive recognition rate of the alternative collaborative combination in the "mountain electromagnetic confrontation" scenario of the target simulation task is 82%, and the task cycle is 23ms.

[0088] The resource utilization rate can be calculated by accumulating the static basic indicators of each simulation model in the alternative collaboration combination and the resource overhead of the simulation device, such as the memory usage of simulation device X is 1.2GB, the CPU utilization of simulation device Y is 40%, and the cross-device communication bandwidth is 15%. The overall resource consumption is calculated, and then the resource utilization rate is comprehensively determined by combining the number of simulation models and simulation devices required for the alternative collaboration combination.

[0089] If it is finally determined that the prediction performance of two alternative collaboration combinations both meet the performance requirements, the first alternative collaboration combination requires the use of three simulation models, while the second alternative collaboration combination only requires the use of two simulation models, then the second alternative collaboration can be selected as the target collaboration combination.

[0090] In some embodiments, a prediction model is trained based on a large number of labeled data sets of performance actually demonstrated by the collaborative combinations in actual tasks. Thus, the prediction performance of the candidate collaborative combinations can be predicted based on the prediction model according to the static basic indicators and scene-sensitive indicators of each candidate simulation model in the candidate collaborative combination.

[0091] It should be understood that if, during the execution of the simulation task, a large amount of resources have been occupied, and the performance parameter set of each simulation model is screened layer by layer according to the simulation model requirements (i.e., pre-screening based on static basic indicators; one round of screening based on the simulation equipment and task scenarios associated with the scene-sensitive indicators of the simulation model, and a second round of screening based on the collaborative combination and task scenarios associated with the scene-sensitive indicators of the simulation model) and a simulation model that matches the performance requirements cannot be selected, then the coordinated screening can be triggered (i.e., pre-screening based on static basic indicators; unified screening based on the simulation equipment, task scenarios, and collaborative combinations associated with the scene-sensitive indicators of the simulation model). Through comprehensive and coordinated planning of the synergistic efficiency of models and equipment, models and models with greater computing power, strong binding screening between simulation models and simulation equipment, and simulation models and simulation models can be achieved. When resources are limited, a combination of "insufficient single performance but complementary combinations" can be selected to meet the overall performance standards.

[0092] In some embodiments, the simulation model requirements include performance requirements, and 105 also includes: based on the performance requirements and the static basic indicators of the simulation model, screening several unused simulation models to obtain several qualified simulation models; generating multiple qualified simulation nodes based on several of the qualified simulation models and several unused simulation devices in the semi-physical simulation resource library, and evaluating the predicted performance of each of the qualified simulation nodes based on the scenario-sensitive indicators of the qualified simulation models and the task scenario of the target simulation task; selecting several qualified simulation nodes whose predicted performance meets the performance requirements; generating several qualified collaborative combinations based on the screened out several qualified simulation nodes that meet the performance requirements, and selecting the simulation model and simulation device in the qualified collaborative combination with the best synergy under the task scenario of the target simulation task as the target simulation model and target simulation device.

[0093] If a qualified simulation node whose predicted performance meets the performance requirements is not selected, for example, a drone model and a visual recognition model are required at this time, only the qualified simulation node corresponding to the drone model that meets the performance requirements is selected, and the qualified simulation node corresponding to the visual recognition model that meets the performance requirements is not selected.

[0094] At this time, based on the performance requirements and the static basic indicators of the simulation model, several unused simulation models are screened to obtain several alternative simulation models; based on several of the alternative simulation models and several of the unused alternative simulation devices in the semi-physical simulation resource library, multiple alternative simulation nodes are generated; based on multiple combinations of the alternative simulation nodes, multiple alternative collaboration combinations are obtained; based on the scene-sensitive indicators of the alternative simulation models, the predicted performance of each of the alternative collaboration combinations in the task scenario of the target simulation task is evaluated; the alternative collaboration combination whose predicted performance meets the performance requirements is selected as the target collaboration combination, and the alternative simulation models and alternative simulation devices in the target collaboration combination are used as the target simulation model and the target simulation device. At this time, the target collaboration combination may contain qualified simulation nodes corresponding to the visual recognition model that does not meet the performance requirements in the layer-by-layer screening, and simulation nodes corresponding to another infrared sensor device, and the two can work together to meet the performance requirements.

[0095] It should be noted that, based on the performance requirements and the static basic indicators of the simulation model, several unused simulation models are screened, and the obtained alternative simulation models or qualified simulation models are essentially the same simulation models, which are named differently in order to distinguish between layer-by-layer screening and overall screening.

[0096] The real-time impact of task scenarios, equipment characteristics and collaboration modes on model performance is quantified through scenario-sensitive indicators. Based on multiple dimensions (simulation performance, resource occupancy), layer-by-layer screening and overall screening mechanisms are used to implement resource scheduling, to cope with performance degradation or complementary effects in a strongly coupled environment, to predict and avoid performance traps, to break through the limitations of single resource performance evaluation, and to optimize the dynamic response capabilities of resource scheduling to complex environments. For example, when resources are sufficient, priority is given to avoiding performance degradation combinations caused by coupling between devices or models; when resources are limited, performance complementarity is achieved through overall planning. For example, the performance of a single model is insufficient, but it can meet the task requirements after being combined with specific equipment or collaboration, thereby maximizing resource utilization, avoiding the waste of resources caused by "information islands" in traditional scheduling, and achieving dynamic adaptation in complex scenarios, improving the efficiency and scalability of the simulation system.

[0097] In some embodiments, by pre-evaluating the predicted performance and resource occupancy of the candidate collaboration combinations, fast switching when resources are scarce is supported, avoiding the rigidity of traditional scheduling caused by fixed resource allocation.

[0098] In some embodiments, the method further includes: loading the target simulation model into a corresponding target simulation device to obtain a number of target semi-physical simulation nodes; in the process of executing the target simulation task, continuously monitoring the task execution quality of a number of target semi-physical simulation nodes; when the task execution quality is lower than a preset quality, obtaining the measured performance data of the target semi-physical simulation node; if the difference between the measured performance data of the target semi-physical simulation node and the performance test data is greater than a preset difference, isolating the simulation data generated by the target semi-physical simulation node.

[0099] Specifically, during the simulation task execution phase, the target simulation model is loaded onto the corresponding target simulation device to form several target semi-physical simulation nodes, such as deploying the communication model to the selected communication simulation device. Subsequently, during the task operation, the task execution quality of each target semi-physical simulation node is monitored and continuously collected in real time. Among them, the task execution quality is used to characterize the actual operation performance of the target semi-physical simulation node, including key parameters such as data accuracy (such as target recognition rate), latency (such as communication response time), and resource utilization (such as CPU occupancy). For example, in the UAV formation simulation, if a target semi-physical simulation node is responsible for path planning, its task execution quality may be composed of indicators such as "path deviation ≤ 2 meters" and "task completion time ≤ 10 seconds".

[0100] When monitoring finds that the task execution quality of a node is lower than the preset quality (such as the path deviation exceeds 3 meters or the task times out), the measured performance data of the target semi-physical simulation node will be obtained, that is, the performance parameters actually collected under the current operating state. Subsequently, the measured performance data is compared with the expected value of the target simulation model in the historical performance test data. If the difference exceeds the preset difference, it is determined that the target semi-physical simulation node is abnormal. At this time, the simulation data generated by the node is immediately isolated, such as cutting off the data flow output to the main system or marking it as invalid data, to prevent abnormal data from affecting the overall simulation results. It should be understood that a series of highly correlated performance test data with the highest similarity to the task scenario type, collaborative combination configuration and simulation equipment model of the target simulation task can be selected to compare the measured performance data.

[0101] Among them, the setting of preset quality and difference threshold can be based on the statistical distribution of historical test data. For example, the upper limit of the 95% confidence interval is taken as the preset quality to ensure that it can respond to abnormalities quickly and avoid misjudgment due to short-term fluctuations. The specific values ​​are not limited here.

[0102] It should be understood that by dynamically comparing the measured data with the expected data, the performance degradation of the simulation model and the simulation equipment caused by environmental fluctuations (such as hardware overload and sudden interference) in real operation can be accurately identified. For example, if a radar model meets the anti-interference capability standard in the test, but the performance drops sharply due to poor heat dissipation of the equipment in actual operation, the difference between the measured data and the test data will trigger isolation to prevent erroneous detection results from contaminating the global simulation.

[0103] In some embodiments, the method also includes: if the difference between the measured performance data and the performance test data of the target semi-physical simulation node is less than the preset difference, based on the scenario-sensitive indicators of the target simulation model, selecting a preferred simulation device from several unused simulation devices in the semi-physical simulation resource library; loading the target simulation model into the preferred simulation device to replace the target simulation device, and updating the target semi-physical simulation node.

[0104] Specifically, during the simulation task execution phase, if the monitoring finds that the difference between the measured performance data of the target semi-physical simulation node and the historical performance test data does not exceed the preset difference, the current combination of the simulation model and the simulation device still has basic performance guarantees. At this time, based on the scenario-sensitive indicators of the model, the preferred simulation device is selected from the unused alternative simulation devices in the resource library. The target simulation model is then migrated from the original target simulation device to the preferred simulation device, and the target semi-physical simulation node is updated, thereby optimizing resource allocation while ensuring task quality.

[0105] Exemplarily, according to the scenario-sensitive index of the current target simulation model and the task scenario and collaborative combination in which it is located, a simulation device that has a gain effect on the model performance of the target simulation model is determined as the preferred simulation model. In other words, associative storage is the basis for quickly locating and eliminating performance fluctuations in resource allocation. The preferred simulation device is a simulation device with better performance in the simulation environment (i.e., task scenario, collaborative combination) in which the current target simulation model is located. Its selection goal is to ensure that the model maintains or improves its performance in a dynamic environment.

[0106] It should be understood that when the task scenario changes suddenly (such as the weather changes from sunny to heavy rain) or equipment failure causes slight performance fluctuations, in order to avoid triggering excessive intervention, alternative combinations can be quickly switched based on the correlation between scene-sensitive indicators, such as enabling a model-device combination with a higher "heavy rain scene recognition rate" without re-testing the entire system. In this way, the synergy potential of simulation models and simulation devices can be identified using scene-sensitive indicators to improve the completion quality of simulation tasks.

[0107] The embodiment of the present application includes steps such as unified management of model resources, dynamic model loading technology, and exception handling mechanism. Among them, the construction of a model resource pool can achieve unified management and scalability of heterogeneous model resources, and improve the reusability and utilization of models; the dynamic model loading technology is to carry out dynamic loading of semi-physical models in simulation, and enhance the flexible deployment of semi-physical simulation models; the exception handling mechanism can realize fault-tolerant control in the semi-physical simulation process, and improve the robustness and stability of the semi-physical simulation system.

[0108] In some embodiments, in order to meet the needs of diverse simulation tasks, it is necessary to uniformly formally describe and functionally encapsulate the simulation models and simulation devices in the semi-physical simulation resource library, map specific resources to logical resources, break down the barriers between collaborative simulation node resources, and achieve unified management of semi-physical simulation resources and comprehensive scheduling during the simulation process, thereby improving the efficiency of resource scheduling and system simulation performance.

[0109] See also Figure 2 , Figure 2 Schematic diagram of a semi-physical simulation resource library package provided by an embodiment of the present application. Figure 2As shown, the simulation resources in the semi-physical simulation resource library include simulation models such as aircraft models, drone models, satellite models, radar models, and simulation devices such as full-quantized node 1, full-quantized node 2, full-quantized node N, lightweight node 1, lightweight node 2, and lightweight node N. Test tasks in different task scenarios are set in advance according to different task requirements in actual application scenarios. For example, the collaborative task requirements include time-space synchronization tasks, transmission tasks, data update tasks, and network transmission tasks; the detection task requirements include search tasks, target indication tasks, and tracking tasks. Based on this, steps S102 to S104 are executed to obtain scene-sensitive indicators at the implementation level, such as effective distance, coverage range, detection accuracy, and bit error rate; static basic indicators at the technical level, such as operating frequency, signal processing capability, and data processing capability. The simulation model is virtualized and packaged from these two levels to obtain a performance parameter set.

[0110] It should be understood that in highly adaptable semi-physical simulation scenarios, the dynamic adaptability and coupling effects of models and devices in different scenarios and collaborative combinations can also capture the implicit dependencies between heterogeneous resources in cross-device and cross-domain scenarios to avoid the formation of "information islands". Furthermore, the real-time impact of task scenarios, equipment characteristics and collaborative modes on model performance is quantified, and resource allocation is matched based on multiple dimensions (simulation performance, resource occupancy) to deal with performance degradation or complementary effects in a strongly coupled environment (such as low-performance models achieving functional compliance through specific equipment combinations). When resources are limited, the utilization rate is improved by optimizing the combination, or the problem is quickly isolated and resources are reconstructed when the task is executed abnormally, which ultimately leads to limited scalability and flexibility of the simulation system, making it difficult to meet the needs of efficient collaboration in complex scenarios.

[0111] See also Figure 3 , Figure 3 1 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device may be a terminal device or a server.

[0112] Exemplarily, the above method can be implemented in the form of a computer program. Figure 3 Runs on the computer device shown.

[0113] like Figure 3 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0114] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any scheduling method for semi-physical simulation resources.

[0115] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0116] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any scheduling method of semi-physical simulation resources.

[0117] The network interface is used for network communication, such as sending assigned tasks.

[0118] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0119] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps: S101, obtaining a semi-physical simulation resource library and test tasks under multiple different task scenarios, wherein the semi-physical simulation resource library includes a plurality of simulation models and a plurality of simulation devices; S102, combining each simulation model with different simulation devices to obtain a plurality of semi-physical simulation nodes, wherein the simulation models or simulation devices of the plurality of semi-physical simulation nodes are different; S103, generating a plurality of collaborative combinations based on the plurality of semi-physical simulation nodes to collaboratively execute a plurality of test tasks, and obtaining performance test data of each simulation model under different simulation devices, different collaborative combinations, and different task scenarios; S104, analyzing the performance test data of each simulation model to generate a performance parameter set for each simulation model, wherein the performance parameter set includes a plurality of scene-sensitive indicators and a plurality of static basic indicators; S105. Based on the task scenario and simulation model requirements of the target simulation task, and according to the simulation model requirements and the performance parameter set of each simulation model, a target simulation model and a target simulation device are selected from the semi-physical simulation resource library to execute the target simulation task.

[0120] Exemplarily, the processor is used to run a computer program stored in the memory, and is also used to implement the steps of the scheduling method for semi-physical simulation resources provided in any embodiment of the present application, which will not be repeated here.

[0121] A computer-readable storage medium is also provided in an embodiment of the present application, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and the processor executes the program instructions to implement the steps of any one of the scheduling methods for semi-physical simulation resources provided in the embodiments of the present application.

[0122] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.

[0123] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A scheduling method for semi-physical simulation resources, characterized in that: The method comprises: S101, obtaining a semi-physical simulation resource library and test tasks under multiple different task scenarios, wherein the semi-physical simulation resource library includes a plurality of simulation models and a plurality of simulation devices; S102, combining each simulation model with different simulation devices to obtain a plurality of semi-physical simulation nodes, wherein the simulation models or simulation devices of the plurality of semi-physical simulation nodes are different; S103, generating a plurality of collaborative combinations based on the plurality of semi-physical simulation nodes to collaboratively execute a plurality of test tasks, and obtaining performance test data of each simulation model under different simulation devices, different collaborative combinations, and different task scenarios; S104, analyzing the performance test data of each simulation model to generate a performance parameter set for each simulation model, wherein the performance parameter set includes a plurality of scene-sensitive indicators and a plurality of static basic indicators; S105. Based on the task scenario and simulation model requirements of the target simulation task, and according to the simulation model requirements and the performance parameter set of each simulation model, a target simulation model and a target simulation device are selected from the semi-physical simulation resource library to execute the target simulation task.

2. The method according to claim 1, characterized in that The performance test data includes several evaluation indicators; S104 includes: Compare multiple values ​​of the same evaluation indicator in the performance test data of each simulation model to determine the fluctuation of the values; When the value fluctuation is greater than a preset fluctuation value, the corresponding evaluation index is determined as the scene-sensitive index, and different values ​​of the scene-sensitive index are associated with the task scene and / or the simulation device and / or the collaborative combination and stored; When the value fluctuation is less than the preset fluctuation value, the corresponding evaluation index is determined as the static basic index.

3. The method according to claim 2, characterized in that The method further comprises: Analyze the changing trends of several values ​​of scene-sensitive indicators, and determine the influence coefficients of task scenarios, simulation equipment, and collaborative combinations on the scene-sensitive indicators; The influencing factor of the scene-sensitive indicator is determined from the task scene, the simulation equipment, and the collaborative combination according to the influencing coefficient, and the influencing factor is associated with the scene-sensitive indicator and stored.

4. The method according to claim 1 or 2, characterized in that: The method further includes: storing the static basic indicator as a fixed value or a range interval, wherein the range interval is determined based on multiple values ​​of the static basic indicator.

5. The method according to claim 1, characterized in that The simulation model requirements include performance requirements, and 105 further includes: Based on the performance requirement and the static basic index of the simulation model, screening a number of unused simulation models to obtain a number of candidate simulation models; Based on a plurality of the candidate simulation models and a plurality of unused candidate simulation devices in the semi-physical simulation resource library, a plurality of candidate simulation nodes are generated; and based on a plurality of combinations of the candidate simulation nodes, a plurality of candidate collaboration combinations are obtained; Based on the scenario-sensitive indicators of the candidate simulation models, evaluating the predicted performance and resource occupancy rate of each of the candidate collaboration combinations in the task scenario of the target simulation task; An alternative collaboration combination whose predicted performance meets the performance requirement and whose resource occupancy rate is the lowest is selected as a target collaboration combination, and an alternative simulation model and an alternative simulation device in the target collaboration combination are selected as the target simulation model and the target simulation device.

6. The method according to claim 1, characterized in that Before S105, the method further includes: obtaining the priorities of the simulation tasks to be executed, and taking at least one simulation task with the highest priority as the target simulation task.

7. The method according to claim 1, characterized in that The method further comprises: Loading the target simulation model into a corresponding target simulation device to obtain a plurality of target semi-physical simulation nodes; During the execution of the target simulation task, continuously monitoring the task execution quality of a plurality of target semi-physical simulation nodes; When the task execution quality is lower than a preset quality, obtaining measured performance data of the target semi-physical simulation node; If the difference between the measured performance data of the target semi-physical simulation node and the performance test data is greater than a preset difference, the simulation data generated by the target semi-physical simulation node is isolated.

8. The method according to claim 7, characterized in that The method further comprises: If the difference between the measured performance data and the performance test data of the target semi-physical simulation node is less than the preset difference, selecting a preferred simulation device from a plurality of unused simulation devices in the semi-physical simulation resource library based on the scene-sensitive index of the target simulation model; The target simulation model is loaded into the preferred simulation device to replace the target simulation device and update the target semi-physical simulation node.

9. A computer device, characterized in that: The device comprises: Memory for storing computer programs; A processor, configured to execute the computer program and implement the method for scheduling semi-physical simulation resources as claimed in any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the method for scheduling semi-physical simulation resources according to any one of claims 1 to 8.

Images