Method for cooperative task scheduling of heterogeneous system based on multi-objective dynamic optimization

Abstract

The present application relates to the technical field of system cooperative task scheduling, and particularly relates to a heterogeneous system cooperative task scheduling method based on multi-target dynamic optimization, which first constructs a task capability model comprising entity jobs, data processing and sensing monitoring units; then performs task decomposition to form a task demand pool and a resource pool; further establishes an efficiency evaluation model of comprehensive processing total efficiency, environment and scene state; based on this, a cooperative scheduling model comprising three-dimensional space representation, dynamic capability fusion, multi-constraint optimization, multi-target decision and closed-loop optimization is constructed by using mixed integer linear programming and heuristic algorithms; finally, resources are configured by a genetic algorithm to generate an optimal scheme based on comprehensive efficiency-cost ratio, drive system interfacing and scene reconstruction, and realize parameter continuous updating.

Description

Technical Field

[0001] This invention relates to the field of system collaborative task scheduling technology, and in particular to a method for collaborative task scheduling of heterogeneous systems based on multi-objective dynamic optimization. Background Technology

[0002] As the complexity of modern systems continues to increase, collaborative operation of heterogeneous systems has become crucial for improving overall operational efficiency. Current technologies typically employ simple hierarchical coordination or loose coupling to allocate tasks between heterogeneous systems, lacking a unified collaborative scheduling framework and precise performance evaluation models.

[0003] Traditional collaborative scheduling methods suffer from the following shortcomings: First, different types of functional units use their own capability evaluation standards, lacking a unified performance representation model, making it difficult to quantify the collaborative effect; second, task allocation and resource scheduling are usually separated, making global optimization impossible; third, scheduling algorithms often adopt centralized computing architectures, which are difficult to cope with the real-time decision-making needs of large-scale collaborative scenarios; in addition, existing methods often focus on optimizing single-objective performance indicators, neglecting the balance between resource consumption and system stability, and lacking effective feedback adjustment mechanisms. Summary of the Invention

[0004] In view of the above problems, the purpose of this invention is to provide a heterogeneous system collaborative task scheduling method based on multi-objective dynamic optimization. This method realizes efficient collaborative task scheduling and dynamic optimization of heterogeneous systems by constructing a unified heterogeneous system task capability model, a dual-pool structure of task demand pool and resource pool, a multi-dimensional performance evaluation model, and a closed-loop optimization system based on cost-effectiveness ratio.

[0005] This invention proposes a cooperative task scheduling method for heterogeneous systems based on multi-objective dynamic optimization, including:

[0006] Step 1: Obtain the target task processing performance parameters of the heterogeneous system and construct the heterogeneous system task capability model. The heterogeneous system task capability model includes a unified mathematical description of the execution capability, control capability and perception capability of three types of functional units: entity operation unit, data processing unit and perception and monitoring unit, for multi-target tasks.

[0007] Step 2: Decompose and organize tasks according to the goals, tasks, and resources of the heterogeneous system, dividing them into a task requirement pool and a task resource pool. The task requirement pool adopts a task tree structure, and the task resource pool includes the task resource status of system nodes, job resources, and environmental information.

[0008] Step 3: Based on the heterogeneous system task capability model, construct a multi-objective task collaborative scheduling system performance evaluation model. The multi-objective task collaborative scheduling system performance evaluation model comprehensively considers the total processing efficiency of the heterogeneous system for all objective tasks, the environmental state, and the scenario state.

[0009] Step 4: Using mixed-integer linear programming and heuristic algorithms, and constrained by the heterogeneous system task capability model and the multi-objective task collaborative scheduling system performance evaluation model, a collaborative scheduling model for collaborative system nodes and multi-objective task processing is constructed using distributed computing technology; the construction of the multi-objective task processing collaborative scheduling model includes:

[0010] A three-dimensional spatial representation model of a collaborative system is constructed, which includes time, space, and efficiency dimensions. The three-dimensional spatial representation model of the collaborative system records the state, spatial location, and mobility of system nodes.

[0011] A dynamic fusion mechanism for heterogeneous system capability models and collaborative scheduling models is constructed. The dynamic fusion mechanism converts the professional capability parameters of different types of units into unified performance evaluation indicators through a capability conversion layer.

[0012] A multi-constraint collaborative optimization algorithm is designed based on a distributed computing framework. The multi-constraint collaborative optimization algorithm decomposes the collaborative scheduling problem into multiple sub-problems and processes them in parallel.

[0013] A multi-objective task decision-making mechanism with adaptive weight dynamic adjustment is constructed. The multi-objective task decision-making mechanism adopts a three-level weight system, and sets weight vectors for task importance, resource scarcity and scenario urgency respectively.

[0014] Design a collaborative scheduling closed-loop optimization system that integrates cost-effectiveness evaluation. The closed-loop optimization system includes six stages: environmental assessment, scheme generation, cost-effectiveness evaluation, scheme optimization, scheme execution, and feedback adjustment.

[0015] Step 5: Using a genetic algorithm, configure the corresponding job resources for the collaborative system nodes to generate the optimal system solution. The optimal system solution is evaluated based on the comprehensive cost-effectiveness ratio index to balance job efficiency and resource consumption.

[0016] Step 6: According to the collaborative system solution, connect the centralized management platform and the functional platform to drive the reconstruction of the scene state and continuously update the multi-objective task processing efficiency parameters of the heterogeneous system.

[0017] Preferably, the heterogeneous system task capability model is a mathematical description of the multi-target task processing efficiency of three types of functional units: physical operation units, data processing units, and perception and monitoring units, wherein:

[0018] Execution capability is quantitatively described through execution effectiveness coefficient, execution accuracy, and target completion coefficient.

[0019] Control capability is quantitatively described by the execution probability, control accuracy, and target completion coefficient under control conditions.

[0020] Perception capability is quantitatively described by execution probability, perception accuracy, and target completion coefficient in the perception state.

[0021] Preferably, the task resource pool is the object domain for collaborative decision-making and task planning, and the elements of the task resource pool include the following three aspects:

[0022] System node task resource status: including the spatial location, current status, speed capability, movement acceleration, reaction time, and task completion progress of the system node;

[0023] Job resources: including platform equipment, resource type, current status, job capacity, and capacity consumption rate;

[0024] Environmental information: This includes the environmental status of heterogeneous system nodes for multi-objective tasks, including meteorological and geographical environments.

[0025] Preferably, the task requirements in the task requirement pool are system task trees. The system task tree includes multiple functional units constituting a task set. The functional units include remote processing units, near-field processing units, data processing units, sensing and monitoring units, and mobile operation units. The system task tree also includes the task sets of the functional units and their task node attributes. The task node attributes include task time nodes, task execution conditions, and task status.

[0026] Preferably, the implementation of the three-dimensional collaborative system spatial representation model includes:

[0027] The system node state description system includes four basic states: active, locked, processing, and protected. The state of each node at each moment is recorded in the state matrix.

[0028] The spatial location representation system uses a three-dimensional coordinate system to record the spatial location of each system node, supporting the monitoring of location changes over continuous time series.

[0029] The mobility parameterization model decomposes the mobility of a node into three key parameters: maximum speed, maximum acceleration, and turning ability, which are used to describe dynamic characteristics.

[0030] The data compression and storage mechanism is designed for large-scale node networks. It selectively stores data based on the importance of nodes, preserving the complete relationships between important nodes and simplifying the records of secondary relationships using probability.

[0031] Preferably, the implementation process of the dynamic fusion mechanism of the heterogeneous system capability model and the cooperative scheduling model includes:

[0032] Capability parameter extraction: Extracting core capability parameters from different functional units;

[0033] Standardization process: unifying various parameters to the same units and range of values;

[0034] Feature mapping: Mapping standardized parameters to a unified performance space based on preset rules;

[0035] Comprehensive calculation: Obtain the synergistic efficiency value after fusion;

[0036] Adaptive weight adjustment: Based on real-time environmental information, the weight ratio of various capability parameters in the fusion calculation is dynamically adjusted.

[0037] Preferably, the task decomposition strategy of the multi-constraint collaborative optimization algorithm includes:

[0038] Domain decomposition: Decomposing a large-scale problem into multiple sub-regional problems according to spatial regions;

[0039] Functional decomposition: Complex problems are broken down into specialized sub-problems according to task type;

[0040] Time decomposition: Decompose a continuous problem into multiple time window problems according to time periods;

[0041] Furthermore, the multi-constraint collaborative optimization algorithm includes two constraint processing mechanisms: hard constraint processing and soft constraint processing. The hard constraints include processing time constraints and resource capacity constraints, while the soft constraints include environmental constraints and collaborative index constraints.

[0042] Preferably, the three-level weighting system of the multi-objective task decision-making mechanism includes:

[0043] Task weighting system: A set of weights based on task importance and priority. Key parameters include priority, task value, and system significance. The adjustment frequency is every time there is a major change in the environment.

[0044] Resource weighting system: A set of weights set according to resource scarcity and system value. Key parameters include resource availability, substitutability, and operational efficiency. The adjustment frequency is based on changes in resource status.

[0045] Environmental weight system: a set of weights based on the rate of change of scene state. Key parameters include the rate of state change, state predictability, and state stability. The adjustment frequency is real-time.

[0046] Preferably, the cost-effectiveness evaluation system of the closed-loop optimization system includes:

[0047] Performance evaluation module: Calculates overall operational performance, including the combined effects of task execution, control, and perception, as well as state stability indicators;

[0048] Cost assessment module: Calculates the cost of task execution, including risk assessment of system node interference and assessment of resource utilization and consumption rate;

[0049] Cost-effectiveness calculation module: Integrates the performance evaluation results and cost evaluation results into a unified comprehensive cost-effectiveness index, which serves as a criterion for judging the merits of a solution.

[0050] Preferably, the anomaly handling mechanism included in the closed-loop optimization system includes:

[0051] State change handling: Set multi-level early warning thresholds, and trigger an emergency replanning process when the state change exceeds the preset threshold;

[0052] Resource failure handling: Establish a dynamic resource replacement mechanism to quickly adjust resource allocation strategies when critical resources fail, ensuring task continuity;

[0053] Task failure handling: Construct a failed task reassignment strategy. When a task fails, replan the task based on the latest environmental assessment results to minimize the cascading impact.

[0054] The present invention has the following beneficial effects:

[0055] 1. It has achieved unified modeling and collaboration of heterogeneous systems, breaking the limitations of different types of functional units working independently, forming a system-level collaborative operation capability, and increasing the coverage of target tasks by approximately 35%;

[0056] 2. By adopting a dual-pool structure of task demand pool and task resource pool, the flexibility of resource allocation and the efficiency of task execution are improved, and the task completion time is reduced by approximately 40%;

[0057] 3. A multi-constraint collaborative optimization algorithm based on a distributed computing framework reduces the computational complexity from O(n³) to O(n²logn), enabling efficient solutions to large-scale collaborative scheduling problems and improving computational speed by 8-15 times;

[0058] 4. The adaptive weight dynamic adjustment mechanism enables the system to proactively perceive and respond to environmental changes, improving environmental sensitivity by approximately 60% and reducing the need for manual intervention by approximately 85%.

[0059] 5. The closed-loop optimization system that integrates cost-effectiveness evaluation balances operational efficiency and resource consumption, improving resource utilization by approximately 32% and system stability by approximately 28%. Attached Figure Description

[0060] Figure 1This is a flowchart of the heterogeneous system collaborative task scheduling method based on multi-objective dynamic optimization according to the present invention. Detailed Implementation

[0061] Please refer to Figure 1 The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0062] Reference Figure 1 This invention provides a method for cooperative task scheduling in heterogeneous systems based on multi-objective dynamic optimization, which includes the following steps:

[0063] Step 1: Obtain the target task processing performance parameters of the heterogeneous system and construct the task capability model of the heterogeneous system.

[0064] In one embodiment of the present invention, it is first necessary to collect performance parameters of different types of functional units, including performance parameters of three types of functional units: physical operation units, data processing units, and sensing and monitoring units. These parameters are usually derived from actual test data, historical operation data, or expert experience values. For example, for physical operation units, parameters such as their effective operating range, execution probability, and reaction time can be obtained; for data processing units, parameters such as their processing range, processing efficiency, and duration can be obtained; and for sensing and monitoring units, parameters such as their sensing range, sensing accuracy, and sensing stability can be obtained.

[0065] Based on the acquired parameters, this invention constructs a unified heterogeneous system task capability model, which provides a unified mathematical description of the execution capability, control capability, and perception capability of three types of functional units. Preferably, the model can be expressed as:

[0066] like , ,

[0067] like , ,

[0068] ,

[0069] in: Functional unit exist Always keep the target task in mind The processing efficiency value; It is a functional unit For the target task The task type (execution objective, control objective, perception objective), with values ​​of 1, 2, and 3; Indicates execution capability. yes The performance coefficient at any given moment. It is a functional unit exist Execution accuracy at any moment It is the completion coefficient of the target task, and its value is in the range of between; Indicates control ability. yes The control effect coefficient at any given time, It is a functional unit exist Precision control of timing; Indicates perceptual ability, yes Perception effect coefficient at any moment It is a functional unit exist Precision of time perception.

[0070] In practical applications, the effectiveness coefficient It is usually dynamically adjusted according to environmental conditions (such as weather conditions and terrain features), with a typical value range of 0.6-1.2; precision parameter Reflecting the technical performance of the functional unit, its typical value range is 0.5-0.95; Target task completion coefficient It is related to the characteristics of the target task. For example, the completion coefficient of dynamic target tasks is usually lower than that of static target tasks, with a typical value range of 0.3-0.9.

[0071] Step 2: Decompose and organize tasks according to the goals, tasks, and resources of the heterogeneous system, dividing them into a task requirement pool and a task resource pool.

[0072] This invention decomposes the collaborative scheduling problem of heterogeneous systems into two parts: a task demand pool and a task resource pool. The task demand pool adopts a task tree structure and contains task sets composed of multiple functional units. In a preferred embodiment of this invention, these functional units include remote processing units, near-field processing units, data processing units, sensing and monitoring units, and mobile operation units. Each functional unit's task set contains corresponding task node attributes, such as task time nodes (start time, end time, critical time points), task execution conditions (preconditions, trigger conditions), and task status (executed, in progress, awaiting allocation, abandoned), etc.

[0073] For example, a comprehensive urban security task might have a hierarchical structure in its system task tree: the top layer represents the urban area targets; the second layer is divided into target monitoring tasks and target processing tasks; and the third layer is further subdivided into specific task nodes such as remote monitoring, mid-range tracking, short-range identification, remote processing, mid-range intervention, and short-range implementation. Each task node has clear time constraints and execution conditions. For instance, the remote processing task must be completed before the target enters a specific area and depends on the successful execution of the remote monitoring task.

[0074] The corresponding task resource pool is the object domain for collaborative decision-making and task planning, encompassing three elements: system node task resource status, operational resources, and environmental information. System node task resource status describes the current state of the system node, such as spatial location (3D coordinates), current state (activated, locked, processing, protected), speed capabilities (maximum speed, cruising speed), acceleration (maximum acceleration, continuous acceleration capability), reaction time (perception-decision-execution link delay), and task completion progress (completion percentage, remaining time). Operational resources include platform equipment (model, quantity), resource type (consumable, non-consumable), current state, operational capabilities, and capability consumption rate. Environmental information includes the environmental status of heterogeneous system nodes for multi-objective tasks, such as weather and geographical environment.

[0075] In practice, the task requirement pool and the task resource pool are stored in a distributed database, supporting real-time updates and efficient queries. The two pools exchange data through a standardized interface to ensure information synchronization and consistency.

[0076] Step 3: Based on the heterogeneous system task capability model, construct a performance evaluation model for the multi-objective task collaborative scheduling system.

[0077] This invention constructs a comprehensive performance evaluation model for a multi-objective task collaborative scheduling system. This model considers three key factors: the overall processing efficiency of the heterogeneous system for all objective tasks, the environmental state, and the scenario state. The mathematical expression of this model can be represented as:

[0078] ,

[0079] ,

[0080] in, It is the overall performance of a heterogeneous system in handling all objectives. The total number of heterogeneous systems; It is a functional unit exist Always on the target The processing efficiency value; For the target task in The environmental state at any given moment; Indicates the functional unit in The degree of influence of the environmental state on the target task at any given moment; This refers to the scene state; Indicates the functional unit in The degree of influence of the scene state on the target task at any given moment.

[0081] In a preferred embodiment of the present invention, the degree of influence $\beta_E^t$ and $\beta_S^t$ are calculated based on a combination of factors, including distance factor, speed factor, and equipment performance factor. For example, the degree of influence can be expressed as:

[0082] ,

[0083] in: These are the weights for distance, speed, and equipment performance, typically set to 0.4, 0.3, and 0.3, respectively. It is a distance factor function, usually an inverse proportional function of distance; It is a velocity factor function, which is proportional to the relative velocity; This is a device performance factor function, related to the device's processing power. The degree of influence of scene state. Taking into account factors such as the complexity of the information environment, the influence of weather conditions, and geographical constraints, it can be expressed as:

[0084] ,

[0085] in: , , These are the weights for the information environment, weather conditions, and geographical environment, with typical values ​​of 0.35, 0.35, and 0.3, respectively. express The information environment complexity at any given time, with a value ranging from 0 to 1; express The degree of influence of weather conditions at any given time, with a value range of 0 to 1; express The geographical environment constraint at any given time, with a value ranging from 0 to 1.

[0086] Through this comprehensive evaluation model, the present invention can fully consider operational efficiency, environmental impact, and scenario factors, providing a scientific basis for subsequent collaborative scheduling optimization.

[0087] Step 4: Using mixed-integer linear programming and heuristic algorithms, and constrained by the heterogeneous system task capability model and the multi-objective task collaborative scheduling system performance evaluation model, a collaborative scheduling model for collaborative system nodes and target processing is constructed using distributed computing technology.

[0088] Step 4 is the core innovation of this invention, which includes five key parts: three-dimensional collaborative system spatial modeling and representation, dynamic fusion mechanism of heterogeneous system capability model and collaborative scheduling model, distributed multi-constraint collaborative optimization algorithm, adaptive weight dynamic adjustment mechanism, and collaborative scheduling closed-loop optimization system with integrated cost-effectiveness evaluation.

[0089] This invention constructs a three-dimensional spatial representation model of a collaborative system, incorporating time, space, and performance dimensions. This model provides a unified representation framework for subsequent collaborative scheduling decisions. In this model, the system node state description system includes four basic states: activation, locking, processing, and protection, forming a state matrix. ,in Represents system nodes exist The state at any given moment.

[0090] Spatial location representation uses a three-dimensional coordinate system to record the spatial location of each system node. It supports monitoring of location changes over continuous time series. The mobility parameterization model decomposes a node's mobility capability into its maximum speed. Maximum acceleration and steering ability The three key parameters, with typical values, are as follows: High-speed moving unit km / h Degrees per second; Medium-speed moving unit km / h , Degrees per second; Ground moving unit km / h , Degrees per second.

[0091] To improve the computational efficiency of large-scale node networks, this invention designs a data compression and storage mechanism based on node importance. This mechanism evaluates node importance using the following formula:

[0092] ,

[0093] in: This represents the importance index of node t; , , These are the weights for capability, location, and relevance, with typical values ​​of 0.4, 0.3, and 0.3, respectively. It is the capability index of a node; It is the value of the node's location system; This refers to the relevance of nodes. An importance index above a certain threshold indicates a high degree of connection. Node relationships with a threshold of 0.7 are fully preserved, while minor relationships below the threshold are recorded using probability simplification, which significantly reduces storage and computational overhead.

[0094] This invention proposes a dynamic fusion mechanism for heterogeneous system capability models and cooperative scheduling models, realizing the organic integration of capabilities of different types of functional units. This fusion mechanism comprises four key steps: capability parameter extraction, standardization processing, feature mapping, and comprehensive calculation.

[0095] Capability parameter extraction involves extracting core capability parameters from different functional units, such as execution range, accuracy, and response time. Standardization processes unify all parameters to the same units and value range [0,1], using the following formula:

[0096] ,

[0097] in: These are the standardized parameter values; These are the original parameter values; and These are the minimum and maximum values ​​for this type of parameter, respectively.

[0098] Feature mapping maps standardized parameters to a unified performance space based on preset rules, constructing a capability transformation matrix. ,in This represents the mapping coefficient from capability type i to effectiveness dimension j. The comprehensive calculation is performed using a fusion function. Computational comprehensive synergistic efficiency .

[0099] This fusion mechanism also includes an adaptive weight adjustment function, which dynamically adjusts the weight ratio of various capability parameters in the fusion calculation based on real-time environmental information. The adjustment formula is as follows:

[0100] ,

[0101] in: for Time-based capability type The weights; This is the smoothing coefficient, typically with a value of 0.7. for The efficiency contribution rate at any given moment. This adaptive mechanism enables the system to flexibly adjust the weights of various capabilities according to the scenario state, maximizing the synergistic effect.

[0102] This invention designs a multi-constraint collaborative optimization algorithm suitable for distributed environments, solving the computational bottleneck of large-scale collaborative scheduling problems. The algorithm employs the MapReduce model, decomposing large-scale problems into parallelizable subproblems, and includes three decomposition strategies: domain decomposition, functional decomposition, and time decomposition.

[0103] Regional decomposition breaks down large-scale problems into multiple sub-regional problems based on spatial regions, such as dividing the entire operational space into multiple relatively independent work areas, with local optimization performed within each area. Functional decomposition breaks down complex problems into specialized sub-problems based on task type, such as breaking down a comprehensive task into monitoring tasks, processing tasks, and intervention tasks. Temporal decomposition breaks down continuous problems into multiple time window problems based on time periods, such as dividing the entire operation process into preparation, execution, and evaluation phases.

[0104] The multi-constraint collaborative optimization algorithm of this invention includes two constraint handling mechanisms: hard constraint handling and soft constraint handling. Hard constraints include handling time-series constraints and resource capacity constraints, which cannot be violated and are directly used as boundary conditions of the MILP model. Soft constraints include environmental constraints and collaborative index constraints, which are added to the objective function through a penalty function and allow conditional violations.

[0105] To ensure consistency among sub-problems, this invention designs a global information sharing pool and a local decision synchronization protocol. The global information sharing pool stores key state variables and constraints, and is updated every 10% of the computational task is completed. The local decision synchronization protocol implements distributed consensus decision-making based on the Paxos consensus algorithm, ensuring that the solutions to each sub-problem are coordinated.

[0106] This distributed optimization algorithm reduces the computational complexity from O(n³) to O(n²logn), significantly improving the computational efficiency of large-scale cooperative scheduling problems. In practical applications, for a problem with 100 nodes, traditional centralized algorithms require approximately 10 minutes of computation time, while this algorithm only requires about 45 seconds, representing a performance improvement of approximately 13 times.

[0107] This invention proposes an adaptive dynamic weight adjustment mechanism based on real-time scenarios, solving the challenge of weight determination in multi-objective optimization. The mechanism employs a three-level weight system, setting weight vectors based on objective importance, resource scarcity, and scenario urgency.

[0108] The target weighting system sets weights based on the importance and priority of the targets. Key parameters include priority level (value 1-5), target value (value 1-10), and system significance (value 1-8). When significant changes occur in the environment, such as the addition of new high-value targets or the increase in the priority level of existing targets, the system will automatically adjust the relevant weights.

[0109] The resource weighting system sets weights based on resource scarcity and system value. Key parameters include resource availability (percentage), substitutability (value 0-1), and operational efficiency (value 1-10). When resource status changes, if a certain type of key resource is consumed too quickly or is lost, the system will adjust the corresponding weight in a timely manner.

[0110] The environmental weighting system sets the weight set based on the scene state change rate. Key parameters include the rate of state change (change per minute), state predictability (value 0-1), and state stability (value 0-1). Environmental weights employ a real-time adjustment mechanism to adapt to the rapidly changing operating environment. The dynamic adjustment of weights uses a gradient-based adaptive adjustment method, with the update formula as follows:

[0111] ,

[0112] in: for The weight vector at each time step; The learning rate is typically set to an initial value of 0.1, and is calculated according to the formula... Attenuation, It is usually taken as 0.01; Let be the gradient of the efficiency function with respect to the weights.

[0113] When conflicts arise among the weights of multiple objectives, the system employs the Analytic Hierarchy Process (AHP) for coordination. First, a judgment matrix is ​​constructed, establishing a relative importance matrix among objectives based on expert knowledge. Then, eigenvectors are calculated to determine the relative weights of each objective. Next, a consistency check is performed to verify the rationality of the judgments. Finally, the calculation results are fused with the dynamic weights to obtain the final weights.

[0114] This adaptive weight adjustment mechanism significantly improves the system's environmental sensitivity and decision-making adaptability. In actual testing, compared to the fixed-weight method, this mechanism reduces the response time to task changes by approximately 75%, improves the ability to detect and respond to minor environmental changes by approximately 60%, and reduces the need for manual intervention by approximately 85%.

[0115] This invention constructs a closed-loop collaborative scheduling optimization system with cost-effectiveness as its core, realizing closed-loop optimization throughout the entire process from evaluation and decision-making to execution. The system comprises three core evaluation modules and six processing stages.

[0116] The three core evaluation modules are: performance evaluation module, cost evaluation module, and cost-effectiveness ratio calculation module. The performance evaluation module calculates overall operational performance, including the combined effects of target execution, control, and perception, as well as state stability indicators. The cost evaluation module calculates the cost of task execution, including risk assessment of system node interference and resource utilization and consumption rate assessment. The cost-effectiveness ratio calculation module integrates performance and cost into a comprehensive cost-effectiveness ratio indicator.

[0117] ,

[0118] in: For overall cost-effectiveness; As a comprehensive performance indicator; The cost to node vitality; The cost of resources and lives.

[0119] The six processing stages of the closed-loop optimization system form a complete OODA loop (Observe-Judgment-Decision-Action): environmental assessment, solution generation, cost-effectiveness assessment, solution optimization, solution execution, and feedback adjustment.

[0120] The environmental assessment phase receives real-time environmental data and historical performance data, extracts environmental features and analyzes trends, and outputs an environmental assessment report and parameter update recommendations. The solution generation phase, based on the environmental assessment results, task requirements, and resource status, uses MILP and heuristic algorithms to generate an initial set of solutions, each containing detailed task allocations and timelines.

[0121] The cost-effectiveness evaluation phase calculates the overall cost-effectiveness of each candidate solution, ranking and filtering them. The solution optimization phase uses a genetic algorithm for iterative optimization to improve the cost-effectiveness of the optimal solution. The solution execution phase generates and issues execution control commands to drive functional units to perform tasks. The feedback and adjustment phase compares the expected results with the actual results, analyzes deviations, and generates parameter adjustment suggestions.

[0122] The system also incorporates a multi-level anomaly handling mechanism, including state mutation handling, resource failure handling, and task failure handling. State mutation handling sets multiple warning thresholds (typically three levels: 15%, 30%, and 50% change), triggering emergency replanning when the state change exceeds the preset threshold. Resource failure handling establishes a dynamic resource replacement mechanism, rapidly adjusting resource allocation strategies when critical resources fail. Task failure handling constructs a failed task redistribution strategy to minimize cascading impacts.

[0123] This closed-loop optimization system balances resource utilization efficiency and operational effectiveness. In practical applications, it increases resource utilization by about 32%, improves system stability by about 28%, reduces decision adjustment frequency by about 45%, and shortens response time to emergencies by about 67%.

[0124] Step 5: Use a genetic algorithm to configure the corresponding job resources for the collaborative system nodes and generate the optimal system solution.

[0125] In step 5 of this invention, a genetic algorithm is used to allocate job resources to the nodes of the collaborative system, generating a system solution with the optimal overall cost-effectiveness. The core idea of ​​the genetic algorithm is to simulate the natural evolutionary process, continuously optimizing the quality of the solution through selection, crossover, and mutation operations.

[0126] In embodiments of the present invention, the chromosome encoding of the genetic algorithm adopts a two-part structure: the first part represents the matching relationship between functional units and the target, and the second part represents the resource allocation scheme. The initial population is typically set to 100-200 individuals, with each individual representing a possible resource allocation scheme. The fitness function directly uses the cost-effectiveness ratio evaluation value; a higher fitness indicates a better scheme quality.

[0127] The selection operation uses a tournament selection method, randomly selecting k individuals (usually k=3 or 5) from the population each time, and choosing the individual with the highest fitness to enter the next generation. The crossover operation uses a two-point crossover method, with the crossover probability typically set to 0.75-0.85. The mutation operation uses uniform mutation, with the mutation probability typically set to 0.05-0.15, decreasing as the number of iterations increases to balance global search and local optimization capabilities.

[0128] The iteration termination condition is usually set at reaching the maximum number of iterations (e.g., 100-200 generations) or no significant improvement in the optimal solution for n consecutive generations (usually n=20). To avoid getting trapped in local optima, the algorithm also introduces an elite retention strategy, ensuring that the best individuals in each generation (usually 5% of the total) directly enter the next generation.

[0129] After optimization using a genetic algorithm, the final optimal system solution includes detailed cooperative system node configurations and resource allocation schemes, achieving an optimal balance between operational efficiency and resource consumption. In actual testing, compared to traditional greedy algorithms or heuristic rules, the solution generated by this method has an average cost-effectiveness improvement of approximately 25%, and exhibits better robustness and adaptability.

[0130] Step 6: Based on the collaborative system solution, connect the centralized management platform and the functional platform to drive the reconstruction of the scenario state and continuously update the target processing performance parameters of the heterogeneous system.

[0131] In the final step of this invention, the system converts the generated optimal collaborative system solution into specific work instructions, distributes them to each functional platform through a centralized management and control platform, and continuously updates the scene status and system parameters based on execution feedback.

[0132] The centralized management platform and functional platforms are connected using standardized command formats and communication protocols to ensure the accuracy and timeliness of command transmission. Preferably, a hierarchical and phased command issuance strategy is adopted, issuing commands for key nodes first, and then issuing other commands gradually in sequence to avoid communication congestion and command conflicts.

[0133] After receiving instructions, the functional platform executes tasks based on its own status and capabilities, and provides real-time feedback on the execution status and results. The centralized management platform dynamically reconstructs the scene status based on this feedback information, including updating information such as target location, status, and environmental assessment.

[0134] The system also continuously updates the target processing performance parameters of the heterogeneous system, including adjusting key parameters such as execution effect coefficient, control effect coefficient, and perception effect coefficient, to adapt to changing operating environments and task requirements. This continuous update mechanism enables the entire system to form a complete closed-loop optimization process, which can continuously improve the quality of decision-making based on actual operational results.

[0135] In practical applications, this closed-loop execution and feedback mechanism significantly improves the system's operational efficiency and adaptability. Compared to the traditional open-loop execution mode, this system can respond to environmental changes more quickly, detect and correct execution deviations in advance, and greatly improve task success rate and resource utilization efficiency.

[0136] In addition to the embodiments described above, the present invention may have other implementations. For example, the three-dimensional collaborative system spatial model may employ different representation methods, such as graph-based network representation; the heterogeneous system capability model may introduce more complex nonlinear relationships; distributed optimization algorithms may be combined with deep reinforcement learning techniques to further improve performance; cost-effectiveness evaluation may consider more dimensional scenario factors, etc. These variations and modifications are all within the scope of protection of the present invention.

[0137] Furthermore, the method of this invention can also be extended to other fields, such as smart city traffic system scheduling and emergency rescue resource coordination. By appropriately adjusting the model parameters and optimization objectives, good collaborative scheduling results can be obtained.

[0138] In summary, the heterogeneous system collaborative task scheduling method based on multi-objective dynamic optimization provided by this invention achieves efficient collaboration and dynamic optimization of heterogeneous systems by constructing a unified heterogeneous system task capability model, an innovative three-dimensional collaborative system spatial representation, an efficient distributed multi-constraint optimization algorithm, an intelligent adaptive weight adjustment mechanism, and a comprehensive cost-effectiveness closed-loop optimization system. This significantly improves the overall operational efficiency and resource utilization efficiency of the system.

[0139] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization, characterized in that, Includes the following steps: Step 1: Obtain the multi-objective task processing performance parameters of the heterogeneous system and construct the heterogeneous system task capability model. The heterogeneous system task capability model includes a unified mathematical description of the execution capability, control capability, and perception capability of three types of functional units: entity operation unit, data processing unit, and perception and monitoring unit, for multi-objective tasks. Step 2: Decompose and organize tasks according to the goals, tasks, and resources of the heterogeneous system, dividing them into a task requirement pool and a task resource pool. The task requirement pool adopts a task tree structure, and the task resource pool includes the system node task resource status, job resources, and environmental information. Step 3: Based on the heterogeneous system task capability model, construct a multi-objective task collaborative scheduling system performance evaluation model. The multi-objective task collaborative scheduling system performance evaluation model comprehensively considers the total processing efficiency of the heterogeneous system for all objective tasks, the environmental state, and the scenario state. Step 4: Using mixed-integer linear programming and heuristic algorithms, and with the heterogeneous system task capability model and the multi-objective task collaborative scheduling system performance evaluation model as constraints, construct a collaborative scheduling model for collaborative system nodes and task processing using distributed computing technology; Step 5: Using a genetic algorithm, configure the corresponding job resources for the collaborative system nodes to generate the optimal system solution. The optimal system solution is evaluated based on the comprehensive cost-effectiveness ratio index to balance job efficiency and resource consumption. Step 6: According to the collaborative system solution, connect the centralized management platform and the functional platform to drive the reconstruction of the scene state and continuously update the multi-objective task processing efficiency parameters of the heterogeneous system.

2. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 1, characterized in that, The construction of the multi-objective task processing collaborative scheduling model includes: A three-dimensional spatial representation model of a collaborative system is constructed, which includes time, space, and efficiency dimensions. The three-dimensional spatial representation model of the collaborative system records the state, spatial location, and mobility of system nodes. A dynamic fusion mechanism for heterogeneous system capability models and collaborative scheduling models is constructed. The dynamic fusion mechanism converts the professional capability parameters of different types of units into unified performance evaluation indicators through a capability conversion layer. A multi-constraint collaborative optimization algorithm is designed based on a distributed computing framework. The multi-constraint collaborative optimization algorithm decomposes the collaborative scheduling problem into multiple sub-problems and processes them in parallel. A multi-objective task decision-making mechanism with adaptive weight dynamic adjustment is constructed. The multi-objective task decision-making mechanism adopts a three-level weight system, and sets weight vectors for task importance, resource scarcity and scenario urgency respectively. Design a collaborative scheduling closed-loop optimization system that integrates cost-effectiveness evaluation. The closed-loop optimization system includes six stages: environmental assessment, scheme generation, cost-effectiveness evaluation, scheme optimization, scheme execution, and feedback adjustment. The heterogeneous system task capability model is a mathematical description of the multi-target task processing efficiency of three types of functional units: physical operation units, data processing units, and perception and monitoring units. Execution capability is quantitatively described through execution effectiveness coefficient, execution accuracy, and target completion coefficient. Control capability is quantitatively described by the execution probability, control accuracy, and target completion coefficient under control conditions. Perception capability is quantitatively described by execution probability, perception accuracy, and target completion coefficient in the perception state.

3. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 2, characterized in that, The task resource pool is the object domain for collaborative decision-making and task planning. The elements of the task resource pool include the following three aspects: System node task resource status: including the spatial location, current status, speed capability, movement acceleration, reaction time, and task completion progress of the system node; Job resources: including platform equipment, resource type, current status, job capacity, and capacity consumption rate; Environmental information: This includes the environmental status of heterogeneous system nodes for multi-objective tasks, including meteorological and geographical environments.

4. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 3, characterized in that, The task requirements in the task requirement pool are a system task tree. The system task tree includes multiple functional units that constitute a task set. The functional units include a remote processing unit, a near-field processing unit, a data processing unit, a sensing and monitoring unit, and a mobile operation unit. The system task tree also includes the task set of the functional units and their task node attributes. The task node attributes include the task time node, the task execution conditions, and the task status.

5. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 4, characterized in that, The implementation methods of the spatial representation model of the three-dimensional collaborative system include: The system node state description system includes four basic states: active, locked, processing, and protected. The state of each node at each moment is recorded in the state matrix. The spatial location representation system uses a three-dimensional coordinate system to record the spatial location of each system node, supporting the monitoring of location changes over continuous time series. The mobility parameterization model decomposes the mobility of a node into three key parameters: maximum speed, maximum acceleration, and turning ability, which are used to describe dynamic characteristics. The data compression and storage mechanism is designed for large-scale node networks. It selectively stores data based on the importance of nodes, preserving the complete relationships between important nodes and simplifying the records of secondary relationships using probability.

6. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 5, characterized in that, The implementation process of the dynamic fusion mechanism of the heterogeneous system capability model and the cooperative scheduling model includes: Capability parameter extraction: Extracting core capability parameters from different functional units; Standardization process: unifying various parameters to the same units and range of values; Feature mapping: Mapping standardized parameters to a unified performance space based on preset rules; Comprehensive calculation: Obtain the synergistic efficiency value after fusion; Adaptive weight adjustment: Based on real-time environmental information, the weight ratio of various capability parameters in the fusion calculation is dynamically adjusted.

7. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 6, characterized in that, The task decomposition strategy of the multi-constraint collaborative optimization algorithm includes: Domain decomposition: Decomposing a large-scale problem into multiple sub-regional problems according to spatial regions; Functional decomposition: Complex problems are broken down into specialized sub-problems according to task type; Time decomposition: Decompose a continuous problem into multiple time window problems according to time periods; Furthermore, the multi-constraint collaborative optimization algorithm includes two constraint processing mechanisms: hard constraint processing and soft constraint processing. The hard constraints include processing time constraints and resource capacity constraints, while the soft constraints include environmental constraints and collaborative index constraints.

8. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 7, characterized in that, The three-level weighting system of the multi-objective task decision-making mechanism includes: Task weighting system: A set of weights based on task importance and priority. Key parameters include priority, task value, and system significance. The adjustment frequency is every time there is a major change in the environment. Resource weighting system: A set of weights set according to resource scarcity and system value. Key parameters include resource availability, substitutability, and operational efficiency. The adjustment frequency is based on changes in resource status. Environmental weight system: a set of weights based on the rate of change of scene state. Key parameters include the rate of state change, state predictability, and state stability. The adjustment frequency is real-time.

9. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 8, characterized in that, The cost-effectiveness evaluation system of the closed-loop optimization system includes: Performance evaluation module: Calculates overall operational performance, including the combined effects of task execution, control, and perception, as well as state stability indicators; Cost assessment module: Calculates the cost of task execution, including risk assessment of system node interference and assessment of resource utilization and consumption rate; Cost-effectiveness calculation module: Integrates the performance evaluation results and cost evaluation results into a unified comprehensive cost-effectiveness index, which serves as a criterion for judging the merits of a solution.

10. The heterogeneous system cooperative task scheduling method based on multi-objective dynamic optimization according to claim 9, characterized in that, The anomaly handling mechanism included in the closed-loop optimization system includes: State change handling: Set multi-level early warning thresholds, and trigger an emergency replanning process when the state change exceeds the preset threshold; Resource failure handling: Establish a dynamic resource replacement mechanism to quickly adjust resource allocation strategies when critical resources fail, ensuring task continuity; Task failure handling: Construct a failed task reassignment strategy. When a task fails, replan the task based on the latest environmental assessment results to minimize the cascading impact.

Images