A distributed intelligent computing resource scheduling optimization method and system

Abstract

The present invention belongs to the field of artificial intelligence computing and scheduling technology, and specifically relates to a distributed intelligent computing resource scheduling optimization method and system, including: real-time acquisition and registration of computing task information sent by each user node, and recording the task information to a blockchain; real-time acquisition and registration of status information of computing resources of each intelligent computing center node, and recording the status information of computing resources to a blockchain; combining the task information of real-time user nodes and the status information of computing resources of the intelligent computing center node, using a reinforcement learning algorithm to infer the optimal scheduling strategy, obtain the optimal scheduling node, and record it to the blockchain; using the optimal scheduling strategy to schedule tasks to the optimal scheduling node, locking resources and triggering task execution, and recording the task execution results to the blockchain; based on the task execution process, constructing and verifying a hash chain, and updating the reputation of the task execution node in real time based on the verification result, forming a closed-loop reward mechanism.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence computing and scheduling technology, and specifically relates to a distributed intelligent computing resource scheduling optimization method and system. Background Art

[0002] In recent years, artificial intelligence (AI) technology has rapidly developed, driving a surge in demand for high-performance computing resources (such as GPUs, TPUs, and FPGAs) for large-scale AI model training and deployment. Computing power has become a critical infrastructure for technological innovation. Furthermore, blockchain technology, leveraging its core advantages of decentralization, transparency, immutability, automated smart contract execution, and trusted record-keeping, has been successfully applied in finance, supply chain management, and data sharing. This provides new insights into trust and collaboration challenges within complex systems and offers potential solutions for optimizing the management of intelligent computing resources.

[0003] Existing intelligent computing resource management approaches suffer from numerous limitations, including low resource utilization efficiency, complex management, limited resource sharing, and a lack of transparency and trustworthiness in task scheduling. Common centralized scheduling methods are prone to single points of failure, potentially causing system paralysis or stagnation. Currently employed resource scheduling strategies, such as static scheduling, round-robin, or load balancing, fail to promptly and accurately respond to the dynamic nature of tasks and resource status, resulting in wasted computing resources and low utilization. Furthermore, existing computing resource scheduling and management processes are often "black-boxed," making it difficult to clearly understand resource usage and the authenticity of execution results across each task. Furthermore, it is difficult to effectively measure and incentivize node contributions and punish misbehavior, potentially leading to a "bad money driving out good money" phenomenon. The integration of blockchain and intelligent computing resource scheduling is currently limited, limited to simple task evidence storage or transaction payments, and lacks comprehensive integration of the entire task lifecycle, including task submission, intelligent matching and scheduling, resource execution monitoring, result verification, and incentive mechanisms. Furthermore, there is a lack of efficient decentralized intelligent scheduling models, making it difficult to meet the real-time, accuracy, and efficiency requirements of intelligent computing resource scheduling. Summary of the Invention

[0004] The present invention provides a distributed intelligent computing resource scheduling optimization method and system. By combining task requirements and the status of intelligent computing center node resources, a reinforcement learning algorithm is used to make intelligent decisions to obtain the optimal scheduling strategy, thereby achieving efficient resource matching and task scheduling optimization, making scheduling decisions more scientific and reasonable, and improving resource utilization and scheduling efficiency. Blockchain technology is used to record information data at each stage to ensure the transparency and traceability of the task execution process, and the tamper-proof nature of the blockchain is used to ensure data authenticity and integrity, and combined with the constructed hash chain for automatic comparison, verification efficiency and accuracy are improved, credibility is improved, and timely error detection and correction are facilitated. At the same time, the task execution results can be used to dynamically adjust the reputation of the task execution node, forming a closed-loop reward mechanism, which helps nodes actively participate in resource scheduling and task execution, and improves system activity and stability.

[0005] A distributed intelligent computing resource scheduling optimization method, comprising:

[0006] Obtain and register computing task information sent by each user node in real time, and record the task information to the blockchain;

[0007] Collect and register the status information of computing resources of each intelligent computing center node in real time, and record the status information of computing resources to the blockchain;

[0008] Combining the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes, the optimal scheduling strategy is inferred using reinforcement learning algorithms to obtain the optimal scheduling node and record it in the blockchain.

[0009] Use the optimal scheduling strategy to schedule tasks to the optimal scheduling node, lock resources and trigger task execution, and record the task execution results to the blockchain;

[0010] Based on the task execution process, a hash chain is constructed and verified, and the reputation of the task execution node is updated in real time based on the verification results, forming a closed-loop reward mechanism.

[0011] By combining task requirements and the status of node resources in the intelligent computing center, and using reinforcement learning algorithms for intelligent decision-making to obtain the optimal scheduling strategy, efficient resource matching and task scheduling optimization are achieved, and scheduling decisions are made more scientific and reasonable, improving resource utilization and scheduling efficiency; blockchain technology is used to record information data at each stage to ensure the transparency and traceability of the task execution process, and the immutability of blockchain is used to ensure data authenticity and integrity, and combined with the constructed hash chain for automated comparison, it improves verification efficiency and accuracy, increases credibility, and helps to timely discover and correct errors; at the same time, the task execution results can be used to dynamically adjust the reputation of the task execution node, forming a closed-loop reward mechanism, which helps nodes actively participate in resource scheduling and task execution, and improves system activity and stability.

[0012] Furthermore, the task information of the real-time user node and the status information of the computing resources of the intelligent computing center node are combined to use the reinforcement learning algorithm to infer the optimal scheduling strategy, obtain the optimal scheduling node, and record it in the blockchain, including:

[0013] Read the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node, build a resource status set of all intelligent computing center nodes, and screen out intelligent computing center nodes with execution capabilities based on the minimum constraint requirements of the current task to build a candidate node set;

[0014] Based on each candidate node in the candidate node set, a state vector of the task-candidate node combination is constructed to represent the static and dynamic scheduling states of the task and the candidate node;

[0015] Based on a given task and a set of candidate nodes, the state vector of the task-candidate node combination is used to calculate the scheduling score of each candidate node using a reinforcement learning algorithm. The candidate node with the highest score is selected as the optimal scheduling node and recorded in the blockchain.

[0016] Furthermore, the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node are read to construct a resource status set of all intelligent computing center nodes, and intelligent computing center nodes with execution capabilities are screened out based on the minimum constraint requirements of the current task to construct a candidate node set, including:

[0017] Read the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node, and build the resource status set of all intelligent computing center nodes;

[0018] Traverse the resource status set of all intelligent computing center nodes and automatically perform logic checks on each intelligent computing center node; the logic checks include reputation checks, resource matching checks, memory available capacity matching checks, type support checks, privacy registration checks, cost budget checks, and time availability checks;

[0019] Based on all intelligent computing center nodes that meet the logical check conditions, a candidate node set is constructed.

[0020] Furthermore, the state vector of the task-candidate node combination is constructed based on each candidate node in the candidate node set to represent the static and dynamic scheduling states of the task and the candidate node, including:

[0021] Perform data cleaning and standardization based on the task parameters and the status parameters of each candidate node in the candidate node combination;

[0022] Extracting processed task parameters and candidate node state parameters to construct a derived feature set of the task and candidate node combination; the derived feature set includes: resource matching degree feature, cost adaptation feature, time adaptation feature, node success rate estimation feature, and type / privacy support binary feature;

[0023] Combine the processed task features, candidate node features, and derived feature sets, perform normalization, and concatenate them to construct the state vector of the task-candidate node combination;

[0024] Based on each task and all its corresponding candidate nodes, the state vectors of the corresponding task-candidate node combination are obtained respectively, and the state vector set of the task-candidate node combination is generated by combination.

[0025] Furthermore, based on a given task and a set of candidate nodes, the state vector of the task-candidate node combination is used to calculate the scheduling score of each candidate node using a reinforcement learning algorithm, and the candidate node with the highest score is selected as the scheduling node and recorded in the blockchain, including:

[0026] Build an intelligent scheduling algorithm model based on deep reinforcement learning;

[0027] Based on the state vector set of the task-candidate node combination formed by the given task and candidate node set, the state vectors of all task-candidate node combinations are trained using the deep reinforcement learning intelligent scheduling algorithm model, the model parameters of the deep reinforcement learning intelligent scheduling algorithm model are output, and a scheduling score result list is obtained;

[0028] Based on the score result list of all candidate nodes, the candidate node corresponding to the maximum score is selected as the optimal scheduling node and recorded in the blockchain.

[0029] Furthermore, it also includes regularly capturing historical information recorded in the blockchain, training and optimizing the intelligent scheduling algorithm model of deep reinforcement learning, including:

[0030] Based on the state vectors of the historical task-candidate node combinations recorded in the blockchain, a historical state vector set is formed;

[0031] Based on the historical state vector set, a scheduling function reward function is constructed, the reward value is calculated, and it is used as a supervision signal;

[0032] Based on the state vector, action, reward value, and next state vector of the historical task-candidate node combination, a standard quadruple is constructed, and a training sample set for reinforcement learning is constructed based on the quadruple;

[0033] The constructed deep reinforcement learning intelligent scheduling algorithm model is trained and optimized using a training sample set, and the optimal deep reinforcement learning intelligent scheduling algorithm model and model parameters are output.

[0034] Furthermore, based on the task execution process, a hash chain is constructed and verified, and the reputation of the task execution node is updated in real time based on the verification results, forming a closed-loop reward mechanism, including:

[0035] Based on the task execution process, log sequences are recorded in real time according to stages;

[0036] Calculate the chain hash of the log sequence in sequence to generate the root hash value of the task log chain, and submit the root hash value to the blockchain by calling the smart contract;

[0037] Reconstruct the hash chain based on the root hash value and all original log sequences in the blockchain;

[0038] Based on the reconstructed hash chain and hash value, a consistency comparison verification is performed to obtain a consistency verification result;

[0039] Based on the consistency verification results, incentive calculations are automatically performed and the reputation of the task execution node is updated to form a closed-loop reward mechanism.

[0040] Furthermore, based on the consistency verification results, the incentive solution is automatically performed and the reputation of the task execution node is updated to form a closed-loop reward mechanism, including:

[0041] Calculate the basic reward value based on the task's resource cost information, the corresponding intelligent computing center node's resource pricing information, and the task execution time;

[0042] Based on the consistency verification results, the log matching rate score of the current task is obtained to determine the credibility of the task execution process;

[0043] Based on the reputation score of the task execution node and the preset weighting factor, the reputation bonus factor of the task execution node is calculated;

[0044] The final reward value is calculated by combining the basic reward value, the credibility of the task execution process, and the reputation bonus factor of the task execution node;

[0045] Based on the credibility of the task execution process, the reputation of the task execution node is adjusted, and the final reward is given to form a closed-loop reward mechanism.

[0046] Furthermore, the credibility of the task execution process is used to adjust the reputation of the task execution node and to make final rewards, including:

[0047] When the credibility of the task execution process is equal to 1, it means that the task execution process is completely credible, the credibility score is improved, and all rewards are issued;

[0048] When the credibility of the task execution process is between 0.8 and 1, it means that the task execution process is basically credible, the credibility score remains basically unchanged or decreases, and rewards are issued proportionally;

[0049] When the credibility of the task execution process is less than 0.8, it means that the task execution process is untrustworthy, the credibility score decreases, and no reward is issued.

[0050] A distributed intelligent computing resource scheduling optimization method system includes a user node set module, an intelligent computing center node set module, a task submission module, a resource status registration module, an intelligent scheduling module, a task execution module, a model optimization training module, a result verification module, and an incentive settlement module;

[0051] The user node set module is used to initiate a computing task to the task submission module;

[0052] The intelligent computing center node set module is used to send the status information of the computing resources of each intelligent computing center node to the resource status registration module;

[0053] The task submission module is used to obtain computing task information initiated by the user node set module in real time and record the task information to the blockchain using the task registration smart contract;

[0054] The resource status registration module is used to collect and register the status information of computing resources of each intelligent computing center node in real time, and record the status information of computing resources to the blockchain using the status registration smart contract;

[0055] The intelligent scheduling module is used to obtain the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes in real time, and obtain the optimal scheduling strategy model output by the model optimization training module;

[0056] The task execution module is used to schedule tasks to the optimal scheduling node using the optimal scheduling strategy, lock resources and trigger task execution, and record the task execution results to the blockchain;

[0057] The model optimization training module is used to use reinforcement learning algorithm reasoning to obtain the optimal scheduling strategy;

[0058] The result verification module is used to build and verify the hash chain based on the task execution process;

[0059] The incentive settlement module is used to update the reputation of the task execution node in real time based on the verification results, forming a closed-loop reward mechanism.

[0060] The beneficial effects of the present invention are:

[0061] The present invention combines the task requirements and the status of the intelligent computing center node resources, uses the reinforcement learning algorithm to make intelligent decisions to obtain the optimal scheduling strategy, realizes efficient resource matching and task scheduling optimization, and makes scheduling decisions more scientific and reasonable, improves resource utilization and scheduling efficiency; uses blockchain technology to record information data at each stage, ensures the transparency and traceability of the task execution process, and uses the non-tamperability of the blockchain to ensure data authenticity and integrity, and combines the constructed hash chain for automated comparison, improves verification efficiency and accuracy, improves credibility, and helps to timely discover and correct errors; at the same time, it can dynamically adjust the reputation of the task execution node using the task execution results, form a closed-loop reward mechanism, help nodes actively participate in resource scheduling and task execution, and improve system activity and stability. By deeply combining the distributed trust mechanism of the blockchain, smart contract technology and artificial intelligence optimization scheduling algorithm, a decentralized intelligent resource scheduling system is constructed, which realizes transparent management of the entire chain in terms of task submission, resource status perception, task intelligent matching, execution process evidence, result verification and incentive settlement, significantly improving computing resource utilization, credibility and scheduling efficiency, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a flow chart of the present invention;

[0063] Figure 2 This is a schematic diagram of the blockchain structure;

[0064] Figure 3 Schematic diagram of the system structure of the present invention;

[0065] Figure 4 A schematic diagram of computer equipment. DETAILED DESCRIPTION

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

[0067] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on this disclosure, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this apparatus and / or practice this method.

[0068] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood in specific situations.

[0069] Example 1

[0070] Figure 1 The method is a distributed intelligent computing resource scheduling optimization method. By combining the task requirements and the status of the intelligent computing center node resources, the reinforcement learning algorithm is used to make intelligent decisions to obtain the optimal scheduling strategy, achieving efficient resource matching and task scheduling optimization, and making scheduling decisions more scientific and reasonable, improving resource utilization and scheduling efficiency; using blockchain technology to record information data at each stage to ensure the transparency and traceability of the task execution process, and using the immutability of the blockchain to ensure data authenticity and integrity, and combining the constructed hash chain for automated comparison, it improves verification efficiency and accuracy, improves credibility, and helps to timely discover and correct errors; at the same time, it can use the task execution results to dynamically adjust the reputation of the task execution node, forming a closed-loop reward mechanism, which helps nodes actively participate in resource scheduling and task execution, and improves system activity and stability. Specifically, it includes the following steps:

[0071] S1: Real-time acquisition and registration of computing task information sent by each user node, and recording the task information to the blockchain;

[0072] In this embodiment, the expression of the task information parameter is:

[0073] ;

[0074] Where, Indicates the The information vector of each task; Indicates the The requirement type of each task, including model training ML, image processing IMG, etc. Indicates the The number of CPU cores for each task; Indicates the The number of GPUs required for each task; Indicates the Memory required for each task; Indicates the The data size of each task; Indicates the The priority of a task ranges from 1 to 10, and the larger the value, the higher the priority; Indicates the The expected start time of each task; Indicates the The maximum tolerable delay of a task; Indicates the The budget cap for each task; Indicates the The maximum execution time of a task; Indicates the The privacy level of each task ranges from 0 to 3. When there is no protection, When it indicates basic isolation, When the encryption is executed, When , it indicates anonymous operation; Indicates the The lower limit of the acceptance node's reputation for a task, with a value range of 0 to 5.

[0075] S2: Collect and register the status information of computing resources of each intelligent computing center node in real time, and record the status information of computing resources to the blockchain;

[0076] In this embodiment, the expression for calculating resource status information parameters is:

[0077] ;

[0078] Where, Indicates the The current resource status vector of each intelligent computing center node, Indicates the total number of intelligent computing center nodes; Indicates the The number of available CPU cores in each intelligent computing center node; Indicates the The number of available GPUs in each intelligent computing center node; Indicates the The total memory capacity of each intelligent computing center node; Indicates the Memory usage ratio of each intelligent computing center node; Indicates the GPU usage rate of each intelligent computing center node; Indicates the Network latency of each intelligent computing center node; Indicates the The node reputation score of each intelligent computing center node ranges from 0 to 5; Indicates the GPU unit price of a smart computing center node; Indicates the The unit price of the CPU of an intelligent computing center node; Indicates the Historical task completion rate of each intelligent computing center node; Indicates the Whether the intelligent computing center node supports remote data sources, its value is 0 or 1, that is, When it indicates that remote data sources are not supported, indicates support for remote data sources; Indicates the The types of tasks that can be processed by each intelligent computing center node; Indicates the The available computing time period of an intelligent computing center node is expressed as: , Indicates the The minimum value of the adjustable time period of each intelligent computing center node; Indicates the The maximum value of the adjustable time period of an intelligent computing center node.

[0079] S3: Combining the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes, the optimal scheduling strategy is inferred using reinforcement learning algorithms to obtain the optimal scheduling node and record it in the blockchain.

[0080] S31: Read the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node, build a resource status set of all intelligent computing center nodes, and screen out intelligent computing center nodes with execution capabilities based on the minimum constraint requirements of the current task to build a candidate node set;

[0081] S311: Read the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node, and build a resource status set of all intelligent computing center nodes;

[0082] The resource status set of all intelligent computing center nodes is:

[0083] ;

[0084] S312: Traverse the resource status set of all intelligent computing center nodes and automatically perform logic checks on each intelligent computing center node;

[0085] The process of automatically performing logic checks includes:

[0086] : Conduct a credit check;

[0087] In this embodiment, the reputation check requires the node reputation Not less than the task tolerance value ,Right now ;

[0088] 2: Perform resource matching check;

[0089] In this embodiment, the resource matching check requires the number of CPU cores required for the task No less than the current number of available CPU cores in the intelligent computing center node , and the number of GPU cores available in the intelligent computing center node Not less than the number of GPUs required for the task ,Right now 、 ;

[0090] 3: Check the available memory capacity.

[0091] In this embodiment, the memory available capacity matching check requires the memory available capacity Not less than the minimum memory requirement for the task ,Right now ;

[0092] Among them, the memory available capacity matching check requires the memory available capacity The calculation expression is:

[0093] ;

[0094] Where, Indicates the total memory capacity of the intelligent computing center node;

[0095] 4: Perform type support check;

[0096] In this embodiment, the type support check requires the task type A set of task types supported by the intelligent computing center node ,Right now ;

[0097] 5: Conduct privacy registration check;

[0098] In this embodiment, the privacy registration check requires the task privacy level Protected and supports privacy execution of intelligent computing center nodes, i.e. ;

[0099] 6: Conduct cost budget check;

[0100] In this embodiment, the cost budget check requires task scheduling and resource cost Not higher than the user's acceptable budget limit ,Right now ;

[0101] Among them, cost budget inspection requires task scheduling and resource cost The calculation expression is:

[0102] ;

[0103] Where, Indicates the Task, Task scheduling and resource costs of each intelligent computing center node;

[0104] 7: Conduct time availability check;

[0105] In this embodiment, the time availability check requires the actual start time of the task Must be located in the schedulable time window of the intelligent computing center node The remaining idle time that the intelligent computing center node can provide is not less than the maximum running time of the task, that is, 、 ;

[0106] Among them, the time availability check requires the actual start time of the task The calculation expression is:

[0107] ;

[0108] Where, Indicates the actual start time of the task; Indicates the current system time;

[0109] S313: Constructing a candidate node set based on all intelligent computing center nodes that meet the logic check conditions;

[0110] In this embodiment, when all logical check conditions are met at the same time, the current intelligent computing center node is included in the candidate node set to form a candidate node set. ;

[0111] Among them, the candidate node set Expressed as:

[0112] ;

[0113] Where, Indicates the total number of candidate nodes.

[0114] S32: Based on each candidate node in the candidate node set, construct a state vector of the task-candidate node combination to represent the static and dynamic scheduling states of the task and the candidate node;

[0115] S321: performing data cleaning and standardization based on the task parameters and the state parameters of each candidate node in the candidate node combination;

[0116] In this embodiment, the data cleaning and standardization process includes:

[0117] S3211: Perform data cleaning based on task parameters, including one-hot encoding of the task types to [1,0,0], [0,1,0], and [0,0,1] for model training (ML), image processing (IMG), and natural language processing (NLP); the expected start time of the task and maximum tolerable delay For non-numeric fields, the original value is retained and scalar normalization is performed.

[0118] S3222: Perform data cleaning based on the status parameters of each candidate node in the candidate node combination, including the memory usage ratio , GPU usage Retention percentage, calculation of remaining memory, candidate node scheduling success probability score, task cost estimate on candidate node, and executable time window length;

[0119] The calculation expression for the remaining memory is:

[0120] ;

[0121] Where, Indicates the The remaining memory of candidate nodes;

[0122] The expression of node scheduling success probability score is:

[0123] ;

[0124] Where, Indicates the The scheduling success probability score of each candidate node;

[0125] The expression for the length of the executable time window is:

[0126] ;

[0127] Where, Indicates the length of the executable time window;

[0128] S3223: Based on the task parameters after data cleaning and the state parameters of each candidate node in the candidate node combination, normalization is performed to make the value range of the parameters be [0, 1].

[0129] S322: extracting the processed task parameters and the state parameters of the candidate nodes, and constructing a derived feature set of the task and candidate node combination;

[0130] In this embodiment, the derived feature set includes resource matching degree feature, cost adaptation feature, time adaptation feature, node success rate estimation feature, type / privacy support binary feature,

[0131] Among them, the resource matching degree characteristics include:

[0132] CPU matching rate , its calculation expression is: ;

[0133] GPU matching rate , its calculation expression is: ;

[0134] Memory matching ratio , its calculation expression is: ;

[0135] Among them, the cost adaptability characteristics include:

[0136] Cost estimation of the task at the candidate node ;

[0137] Budget matching , its calculation expression is: ;

[0138] Among them, the time adaptability features include:

[0139] Executable time window length , its calculation expression is: ;

[0140] Time margin ratio , its calculation expression is: ;

[0141] Among them, the node success rate estimation features include:

[0142] Weighted Success Factor , its calculation expression is: ;

[0143] Among them, type / privacy supports binary features, including:

[0144] Does the task type match? , whose expression is:

[0145] ;

[0146] Privacy support matches , its calculation expression is:

[0147] ;

[0148] S323: combining the processed task features, candidate node features, and derived feature sets, performing normalization processing, and concatenating and constructing a state vector of the task-candidate node combination;

[0149] Combine the processed task features, candidate node features, and derived feature sets to construct the state vector of the task-candidate node combination ;

[0150] ;

[0151] Where, Represents task characteristics; Represents the candidate node features; Indicates derived features;

[0152] The state vectors of the task-candidate node combinations shown in Table 1 Contents:

[0153] Table 1 Contents of the state vector of the task-candidate node combination

[0154] Classification Feature Field Task features Resource requirements (CPU / GPU / MEM), budget, runtime, privacy level, task type (one-hot) Candidate node features Available resources, unit price, reputation value, support type, and whether privacy enforcement is supported Derived features (match features) Cost estimation, time adaptation, resource margin, success rate prediction, waiting delay, load information

[0155] S324: Based on each task and all corresponding candidate nodes, the state vectors of the corresponding task-candidate node combination are obtained respectively, and the state vector set of the task-candidate node combination is generated by combination.

[0156] Based on each task to be scheduled after pre-screening and the corresponding candidate node set , repeat S321~S323, obtain the state vector of the corresponding task-candidate node combination, and encode it into a structured numerical vector , and combine to generate a state vector set of task-candidate node combinations ;

[0157] Among them, the state vector set of the generated task-candidate node combination Expressed as:

[0158] ;

[0159] Where, , express dimensional real number space.

[0160] S33: Based on a given task and a set of candidate nodes, the state vector of the task-candidate node combination is used to calculate the scheduling score of each candidate node using a reinforcement learning algorithm. The candidate node with the highest score is selected as the optimal scheduling node and recorded in the blockchain.

[0161] S331: Constructing an intelligent scheduling algorithm model based on deep reinforcement learning;

[0162] Among them, the expression of the constructed deep reinforcement learning intelligent scheduling algorithm model is:

[0163] ;

[0164] Where, Indicates that the Assign tasks to The long-term scheduling benefit estimation that can be obtained after candidate nodes; Represents the weight matrix and bias terms of all layers; represents the weight matrix of the first hidden layer; represents the bias vector of the first hidden layer; represents the weight matrix of the second hidden layer; represents the bias vector of the second hidden layer; represents the weight matrix of the output layer; represents the bias vector of the output layer; Represents the activation function, whose expression is: ;

[0165] S332: Based on the state vector set of the task-candidate node combination formed by the given task and the candidate node set, the state vectors of all task-candidate node combinations are trained using the deep reinforcement learning intelligent scheduling algorithm model, model parameters of the deep reinforcement learning intelligent scheduling algorithm model are output, and each task-candidate node combination is scored based on the trained deep reinforcement learning intelligent scheduling algorithm model to obtain a scheduling score result list;

[0166] Among them, the scheduling score result list is expressed as:

[0167] ;

[0168] Where, Represents a given task Down, The scheduling score results of candidate nodes;

[0169] S333: Based on the score result list of all candidate nodes, select the candidate node corresponding to the maximum score, use it as the optimal scheduling node, and record it in the blockchain;

[0170] Among them, based on the score result list of all candidate nodes , the optimal scheduling node number is selected as:

[0171] ;

[0172] Where, represents the optimal scheduling node;

[0173] S4: Schedule tasks to the optimal scheduling node using the optimal scheduling strategy, lock resources and trigger task execution, and record the task execution results to the blockchain;

[0174] In this embodiment, the task and optimal scheduling nodes As input, the scheduling registration smart contract is used to schedule and the current timestamp Record to the blockchain, and then call the resource locking contract according to the scheduling The corresponding resources are locked. When the lock is successful, the task information is sent to the optimal scheduling node through broadcasting. , trigger the task execution, and the task execution result and the timestamp segment from the start to the end of the task Recorded to blockchain, Indicates the task start timestamp, Indicates the task end timestamp.

[0175] In this embodiment, the process also includes periodically capturing historical information recorded on the blockchain and training and optimizing the intelligent scheduling algorithm model based on deep reinforcement learning, including:

[0176] A1: State vector of the task-candidate node combination based on the history recorded in the blockchain , forming a set of historical state vectors ;

[0177] A2: Based on the historical state vector set , construct the scheduling function reward function, calculate the reward value, and use it as a supervision signal;

[0178] The calculation expression of the reward value is:

[0179] ;

[0180] Where, Indicates that the Assign tasks to The reward value obtained by the candidate node; represents the success rate weight factor; represents the cost weight factor; represents the weight factor of the delay penalty term; Indicates a task Scheduled to the optimal scheduling node The estimated waiting time in the task queue is the difference between the timestamp sent to the receiver according to the schedule in the historical system log. Its expression is: ; represents the load penalty factor; Represents the optimal scheduling node The current resource usage is expressed as: ;

[0181] A3: Based on the state vector, action, reward value, and next state vector of the historical task-candidate node combination, a standard quadruple is constructed, and a training sample set for reinforcement learning is constructed based on the quadruple. ;

[0182] Among them, the standard four-tuple is expressed as:

[0183] ;

[0184] Where, Indicates the current The combined feature vector of the task and the state vector of the node at the moment; express Moment reward value; Indicates the next The combined feature vector of the task and the state vector of the node at the moment;

[0185] A4: Using a training sample set Train and optimize the constructed deep reinforcement learning intelligent scheduling algorithm model, and output the optimal deep reinforcement learning intelligent scheduling algorithm model and model parameters;

[0186] Among them, the loss function of the intelligent scheduling algorithm model using deep reinforcement learning is:

[0187] ;

[0188] Where, Represents the loss function value of the intelligent scheduling algorithm model of deep reinforcement learning; Indicates that the status Take action When the model parameters are The action value function predicted by the intelligent scheduling algorithm model of deep reinforcement learning; Indicates reward; Indicates the next state Next, select the action that maximizes the action value function The corresponding action value;

[0189] S5: Based on the task execution process, a hash chain is constructed and verified, and the reputation of the task execution node is updated in real time based on the verification results, forming a closed-loop reward mechanism.

[0190] S51: Based on the task execution process, log sequence is recorded in real time according to the stage ;

[0191] Among them, the log sequence , Indicates that the task has started executing. Indicates that the task calculation processing is completed;

[0192] S52: Calculate the chain hash of the log sequence in sequence to generate the root hash value of the task log chain , and use the call smart contract to set the root hash value Submit to blockchain;

[0193] Among them, the expression of chain hash is:

[0194] ;

[0195] Where, Indicates the The hash value calculated from the task log sequence; represents a hash function;

[0196] S53: Reconstruct the hash chain based on the root hash value and all original log sequences in the blockchain;

[0197] Among them, the expression of the reconstructed hash chain is:

[0198] ;

[0199] Where, Represents the reconstructed hash chain; Represents the hash chain reconstruction function;

[0200] S54: Perform consistency comparison verification based on the reconstructed hash chain and hash value to obtain a consistency verification result;

[0201] Among them, based on the reconstructed hash chain and hash value, consistency comparison verification is performed, and its expression is:

[0202] ;

[0203] Where, Indicates the credibility of the task execution process;

[0204] In this embodiment, when , it means the consistency verification is passed, otherwise the verification fails.

[0205] S55: Based on the consistency verification results, incentive calculation is automatically performed and the reputation of the task execution node is updated to form a closed-loop reward mechanism;

[0206] S551: Calculate the basic reward value based on the resource cost information of the task, the resource pricing information of the corresponding intelligent computing center node, and the task execution time;

[0207] In this embodiment, the task The resource cost information includes CPU information, GPU information and memory requirement information, which is expressed as: ; Corresponding intelligent computing nodes The resource pricing is expressed as: ;

[0208] The calculation expression of the basic reward value is:

[0209] ;

[0210] Where, Indicates the Task, The basic reward value of each intelligent computing center node;

[0211] S552: Based on the consistency verification result, the log matching rate score of the current task is obtained to determine the credibility of the task execution process;

[0212] In this embodiment, based on the consistency verification result, the log matching rate score of the current task is obtained to determine the credibility of the task execution process. , its value range is 0~1;

[0213] Among them, the credibility of the task execution process The calculation expression is:

[0214] ;

[0215] Where, Indicates the credibility of the task execution process; Indicates the number of log entries that pass the hash comparison; Indicates the expected number of commit logs;

[0216] S553: ​​Calculate the reputation bonus factor of the task execution node based on the reputation score of the task execution node and the preset weighting factor;

[0217] In this embodiment, the reputation score of the task execution node is based on , combined with the preset weighting factors , the expression for calculating the reputation bonus factor of the task execution node is:

[0218] ;

[0219] Where, Represents the reputation bonus factor of the task execution node; Indicates the system preset weighting factor;

[0220] S554: Calculate the theoretical reward value by combining the basic reward value, the credibility of the task execution process, and the reputation bonus factor of the task execution node;

[0221] Among them, the calculation expression of the theoretical reward value is:

[0222] ;

[0223] Where, Indicates the theoretical final reward value;

[0224] S555: Based on the credibility of the task execution process, the reputation of the task execution node is adjusted and the final reward is given to form a closed-loop reward mechanism;

[0225] S5551: When the task execution process is reliable Equal to 1.0, that is When , it means that the task execution process is completely trustworthy, the credit score is improved, and all rewards are issued;

[0226] Among them, when , that is, when the task execution process is completely trustworthy, the current credit score is adjusted to:

[0227] ;

[0228] Where, Indicates the updated reputation score of the task execution node; Indicates the reputation score value of the task execution node before the update; Indicates the improved reputation score value, and its value range is [0.01, 0.05];

[0229] At this time, the reward coefficient is issued , that is, the final reward value is:

[0230] ;

[0231] Where, Indicates the actual final reward value; Indicates the reward coefficient;

[0232] S5552: When the task execution process is reliable is 0.8~1.0, that is When , it means that the task execution process is basically credible but there are omissions. The credit score remains basically unchanged or slightly decreases, and rewards are issued proportionally.

[0233] Among them, when , that is, when the task execution process is basically credible, the current credit score is adjusted to:

[0234] ;

[0235] Where, Represents a basically unchanged reputation score value, that is, ;

[0236] At this time, the reward coefficient is issued The expression is:

[0237] ;

[0238] That is, the calculation expression of the final reward value is:

[0239] ;

[0240] S5553: When the task execution process is reliable Not more than 0.8, that is If the task execution process is not credible or there is suspicion of fraud, the credit score will drop or a penalty period will be imposed, and no reward will be issued;

[0241] Among them, when , that is, when the task execution process is untrustworthy, the current credit score is adjusted to:

[0242] ;

[0243] Where, Indicates the credit score value that has decreased, and its value range is [0.05, 0.1];

[0244] At this time, the reward coefficient is issued , that is, the final reward value is:

[0245] .

[0246] Table 2 shows the credit score adjustment and reward issuance rule settings based on the credibility of the task execution process.

[0247] Table 2 Credibility score adjustment and reward issuance rule setting based on the credibility of task execution process

[0248] ;

[0249] In this embodiment, if Figure 2 Shown is the architecture diagram of the blockchain.

[0250] Example 2

[0251] Based on the same technical concept, such as Figure 3 As shown, this embodiment also provides a distributed intelligent computing resource scheduling and optimization system, including a user node set module, an intelligent computing center node set module, a task submission module, a resource status registration module, an intelligent scheduling module, a task execution module, a model optimization training module, a result verification module, and an incentive settlement module.

[0252] Specifically, the user node set module is used to initiate computing tasks to the task submission module;

[0253] Specifically, the intelligent computing center node set module is used to send the status information of the computing resources of each intelligent computing center node to the resource status registration module;

[0254] Specifically, the task submission module is used to obtain computing task information initiated by the user node set module in real time, and use the task registration smart contract to record the task information to the blockchain;

[0255] Specifically, the resource status registration module is used to collect and register the status information of computing resources of each intelligent computing center node in real time, and use the status registration smart contract to record the status information of computing resources to the blockchain;

[0256] Specifically, the intelligent scheduling module is used to obtain the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes in real time, as well as the optimal scheduling strategy model output by the model optimization training module;

[0257] Specifically, the task execution module is used to schedule tasks to the optimal scheduling node using the optimal scheduling strategy, lock resources and trigger task execution, and record the task execution results to the blockchain;

[0258] Specifically, the model optimization training module is used to obtain the optimal scheduling strategy using reinforcement learning algorithm reasoning;

[0259] Specifically, the result verification module is used to build and verify the hash chain based on the task execution process;

[0260] Specifically, the incentive settlement module is used to update the reputation of task execution nodes in real time based on verification results, forming a closed-loop reward mechanism.

[0261] Example 3

[0262] Based on the same technical concept, the embodiment of the present application also provides a computer device, including a memory 1 and a processor 2, such as Figure 4 As shown, the memory 1 stores a computer program, and the processor 2 implements any of the above methods when executing the computer program.

[0263] Memory 1 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, memory 1 can be an internal storage unit of the distributed intelligent computing resource scheduling and optimization system, such as a hard disk. In other embodiments, memory 1 can also be an external storage device of the distributed intelligent computing resource scheduling and optimization system, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. Furthermore, memory 1 can include both an internal storage unit of the distributed intelligent computing resource scheduling and optimization system and an external storage device. Memory 1 can be used not only to store application software and various types of data installed in the distributed intelligent computing resource scheduling and optimization system, such as the code of the distributed intelligent computing resource scheduling and optimization system program, but can also be used to temporarily store data that has been output or is about to be output.

[0264] In some embodiments, the processor 2 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in the memory 1 or process data, such as executing a distributed intelligent resource scheduling optimization system program.

[0265] The disclosed embodiments of the present invention further provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method described in the above method embodiments. The storage medium may be a volatile or non-volatile computer-readable storage medium.

[0266] The computer program product of the application page content refresh method provided in the disclosed embodiment of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the steps of the method described in the above method embodiment. For details, please refer to the above method embodiment and will not be repeated here.

[0267] The disclosed embodiments of the present invention further provide a computer program that, when executed by a processor, implements any of the methods of the aforementioned embodiments. This computer program product can be implemented in hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is embodied as a computer storage medium. In another alternative embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK).

[0268] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0269] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" is at least two.

[0270] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0271] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0272] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0273] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0274] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0275] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0276] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A distributed intelligent computing resource scheduling optimization method, characterized in that: include: Obtain and register computing task information sent by each user node in real time, and record the task information to the blockchain; Collect and register the status information of computing resources of each intelligent computing center node in real time, and record the status information of computing resources to the blockchain; Combining the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes, the optimal scheduling strategy is inferred using reinforcement learning algorithms to obtain the optimal scheduling node and record it in the blockchain. Use the optimal scheduling strategy to schedule tasks to the optimal scheduling node, lock resources and trigger task execution, and record the task execution results to the blockchain; Based on the task execution process, a hash chain is constructed and verified, and the reputation of the task execution node is updated in real time based on the verification results, forming a closed-loop reward mechanism; Based on the task execution process, the hash chain is constructed and verified, and the reputation of the task execution node is updated in real time based on the verification results, forming a closed-loop reward mechanism, including: Based on the task execution process, log sequences are recorded in real time according to stages; Calculate the chain hash of the log sequence in sequence to generate the root hash value of the task log chain, and submit the root hash value to the blockchain by calling the smart contract; Reconstruct the hash chain based on the root hash value and all original log sequences in the blockchain; Based on the reconstructed hash chain and the root hash value, a consistency comparison is performed to obtain a consistency verification result; Based on the consistency verification results, incentive calculation is automatically performed and the reputation of the task execution node is updated to form a closed-loop reward mechanism; Based on the consistency verification results, the incentive solution is automatically calculated and the reputation of the task execution node is updated to form a closed-loop reward mechanism, including: Calculate the basic reward value based on the task's resource cost information, the corresponding intelligent computing center node's resource pricing information, and the task execution time; Based on the consistency verification results, the log matching rate score of the current task is obtained to determine the credibility of the task execution process; Based on the reputation score of the task execution node and the preset weighting factor, the reputation bonus factor of the task execution node is calculated; The final reward value is calculated by combining the basic reward value, the credibility of the task execution process, and the reputation bonus factor of the task execution node; Based on the credibility of the task execution process, the reputation of the task execution node is adjusted, and the final reward is given to form a closed-loop reward mechanism.

2. A distributed intelligent computing resource scheduling optimization method according to claim 1, characterized in that: The method combines the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes, uses reinforcement learning algorithm to infer the optimal scheduling strategy, obtains the optimal scheduling node, and records it in the blockchain, including: Read the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node, build a resource status set of all intelligent computing center nodes, and screen out intelligent computing center nodes with execution capabilities based on the minimum constraint requirements of the current task to build a candidate node set; Based on each candidate node in the candidate node set, a state vector of the task-candidate node combination is constructed to represent the static and dynamic scheduling states of the task and the candidate node; Based on a given task and a set of candidate nodes, the state vector of the task-candidate node combination is used to calculate the scheduling score of each candidate node using a reinforcement learning algorithm. The candidate node with the highest score is selected as the optimal scheduling node and recorded in the blockchain.

3. A distributed intelligent computing resource scheduling optimization method according to claim 2, characterized in that: The parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node are read to build a resource status set of all intelligent computing center nodes, and intelligent computing center nodes with execution capabilities are screened out based on the minimum constraint requirements of the current task to build a candidate node set, including: Read the parameter vector of the current task information and the status information of the computing resources of the intelligent computing center node, and build the resource status set of all intelligent computing center nodes; Traverse the resource status set of all intelligent computing center nodes and automatically perform logic checks on each intelligent computing center node; the logic checks include reputation checks, resource matching checks, memory available capacity matching checks, type support checks, privacy registration checks, cost budget checks, and time availability checks; Based on all intelligent computing center nodes that meet the logical check conditions, a candidate node set is constructed.

4. A distributed intelligent computing resource scheduling optimization method according to claim 2, characterized in that: The state vector of the task-candidate node combination is constructed based on each candidate node in the candidate node set to represent the static and dynamic scheduling states of the task and the candidate node, including: Perform data cleaning and standardization based on the task parameters and the status parameters of each candidate node in the candidate node combination; Extracting processed task parameters and candidate node state parameters to construct a derived feature set of the task and candidate node combination; the derived feature set includes: resource matching degree feature, cost adaptation feature, time adaptation feature, node success rate estimation feature, and type / privacy support binary feature; Combine the processed task features, candidate node features, and derived feature sets, perform normalization, and concatenate them to construct the state vector of the task-candidate node combination; Based on each task and all its corresponding candidate nodes, the state vectors of the corresponding task-candidate node combination are obtained respectively, and the state vector set of the task-candidate node combination is generated by combination.

5. A distributed intelligent computing resource scheduling optimization method according to claim 4, characterized in that: Based on a given task and a set of candidate nodes, the state vector of the task-candidate node combination is used to calculate the scheduling score of each candidate node using a reinforcement learning algorithm, and the candidate node with the highest score is selected as the scheduling node and recorded in the blockchain, including: Build an intelligent scheduling algorithm model based on deep reinforcement learning; Based on the state vector set of the task-candidate node combination formed by the given task and candidate node set, the state vectors of all task-candidate node combinations are trained using the deep reinforcement learning intelligent scheduling algorithm model, the model parameters of the deep reinforcement learning intelligent scheduling algorithm model are output, and a scheduling score result list is obtained; Based on the score result list of all candidate nodes, the candidate node corresponding to the maximum score is selected as the optimal scheduling node and recorded in the blockchain.

6. A distributed intelligent computing resource scheduling optimization method according to claim 5, characterized in that: It also includes regularly capturing historical information recorded on the blockchain, training and optimizing the intelligent scheduling algorithm model based on deep reinforcement learning, including: Based on the state vectors of the historical task-candidate node combinations recorded in the blockchain, a historical state vector set is formed; Based on the historical state vector set, a scheduling function reward function is constructed, the reward value is calculated, and it is used as a supervision signal; Based on the state vector, action, reward value, and next state vector of the historical task-candidate node combination, a standard quadruple is constructed, and a training sample set for reinforcement learning is constructed based on the quadruple; The constructed deep reinforcement learning intelligent scheduling algorithm model is trained and optimized using a training sample set, and the optimal deep reinforcement learning intelligent scheduling algorithm model and model parameters are output.

7. A distributed intelligent computing resource scheduling optimization method according to claim 1, characterized in that: The credibility of the task execution process is used to adjust the reputation of the task execution node and to make final rewards, including: When the credibility of the task execution process is equal to 1, it means that the task execution process is completely credible, the credibility score is improved, and all rewards are issued; When the credibility of the task execution process is between 0.8 and 1, it means that the task execution process is basically credible, the credibility score remains basically unchanged or decreases, and rewards are issued proportionally; When the credibility of the task execution process is less than 0.8, it means that the task execution process is untrustworthy, the credibility score decreases, and no reward is issued.

8. A system for the distributed intelligent computing resource scheduling optimization method according to claim 1, characterized in that: It includes user node set module, intelligent computing center node set module, task submission module, resource status registration module, intelligent scheduling module, task execution module, model optimization training module, result verification module, and incentive settlement module; The user node set module is used to initiate a computing task to the task submission module; The intelligent computing center node set module is used to send the status information of the computing resources of each intelligent computing center node to the resource status registration module; The task submission module is used to obtain computing task information initiated by the user node set module in real time and record the task information to the blockchain using the task registration smart contract; The resource status registration module is used to collect and register the status information of the computing resources of each intelligent computing center node in real time, and record the status information of the computing resources to the blockchain using the status registration smart contract; The intelligent scheduling module is used to obtain the task information of real-time user nodes and the status information of computing resources of intelligent computing center nodes in real time, and obtain the optimal scheduling strategy model output by the model optimization training module; The task execution module is used to schedule tasks to the optimal scheduling node using the optimal scheduling strategy, lock resources and trigger task execution, and record the task execution results to the blockchain; The model optimization training module is used to use reinforcement learning algorithm reasoning to obtain the optimal scheduling strategy; The result verification module is used to build and verify the hash chain based on the task execution process; The incentive settlement module is used to update the reputation of the task execution node in real time based on the verification results, forming a closed-loop reward mechanism.

Images