Intelligent laboratory management system and method of fusion graph machine learning

Abstract

The invention discloses an intelligent laboratory management system for fusion graph machine learning, and the system comprises a data layer which is configured to store data related to laboratory management and comprises a business database, a process database, an instrument state database and a knowledge graph database; the AI middle table layer is connected with the data layer and comprises a graph machine learning engine and a reinforcement learning engine; the application layer is connected with the AI middle table layer and is used for calling the service of the AI middle table layer to realize a laboratory service function; and the user layer is connected with the application layer and is used for providing an interactive interface for users of different roles. According to the invention, the AI capability is embedded into the laboratory management process by constructing the machine learning and reinforcement learning of the intelligent platform fusion graph, so that the multi-source data integration and mining are realized, and the information isolated island is effectively broken; resource configuration is optimized through dynamic scheduling, and the operation efficiency is improved; by means of active risk prediction and propagation analysis, early risk identification and intelligent management and control are realized, and automation, intelligence and optimization levels of laboratory management are further improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent laboratory information management technology, and more specifically, to an intelligent laboratory management system and method that integrates graph machine learning. Background Technology

[0002] With the increasing frequency and complexity of scientific research activities, chemistry laboratories, as one of the core locations for research and teaching, face severe challenges in terms of management efficiency and intelligence. Traditional laboratory information management systems primarily achieve electronic recording of business processes, but they have significant shortcomings in data integration and intelligent decision-making. Existing systems generally suffer from information silos, with inconsistent data formats in stages such as quotations, contracts, sampling, testing, and reporting, relying on manual transmission and repetitive entry, resulting in low efficiency and a high risk of errors. Meanwhile, resource scheduling and task allocation are mostly based on fixed rules or human experience, lacking dynamic adaptability and struggling to cope with emergencies such as instrument malfunctions and urgent samples, easily leading to uneven resource utilization and project delays. Furthermore, existing technologies lack proactive and intelligent early warning mechanisms for potential risks in the experimental process, such as expired samples, abnormal results, and poor instrument condition; report generation is cumbersome, and management decisions lack in-depth analysis support based on historical data.

[0003] To address these challenges, existing technologies have attempted to incorporate the Internet of Things (IoT) and automated process management, such as through system architecture to achieve data collection and task matching. Other solutions propose using artificial intelligence (AI) to build integrated collaborative platforms for information extraction and risk prediction. However, these existing technologies still fall short in deep fusion and correlation mining of multi-source data, and bottlenecks exist in platform connectivity and data transmission efficiency, failing to fundamentally solve the core need for intelligent management of the entire laboratory process. Therefore, there is an urgent need for a new generation of laboratory management systems that can deeply integrate advanced AI algorithms to achieve intelligent data integration, proactive risk warning, and dynamic resource optimization. Summary of the Invention

[0004] To address the aforementioned technical problems in related technologies, this invention proposes an intelligent laboratory management system and method that integrates graph machine learning, which can overcome the above-mentioned shortcomings of the prior art.

[0005] To achieve the above-mentioned technical objectives, the technical solution of the present invention is implemented as follows: An intelligent laboratory management system that integrates graph machine learning; This intelligent laboratory management system, which integrates graph machine learning, includes a data layer, an AI platform layer, an application layer, and a user layer. The data layer is configured to store laboratory management-related data, including a business database, a process database, an instrument status database, and a knowledge graph library. The AI ​​platform layer is connected to the data layer and includes a graph machine learning engine and a reinforcement learning engine. The application layer is connected to the AI ​​platform layer and is used to call the services of the AI ​​platform layer to realize laboratory business functions. The user layer is connected to the application layer and is used to provide interactive interfaces for users with different roles.

[0006] Furthermore, the knowledge graph database is constructed using a graph database, and the types of nodes stored include samples, projects, inspectors, instruments, customers, testing methods, and testing indicators; the types of edges stored include at least one of "belongs to", "responsible for testing", "use", "requires qualification", "abnormal result", "failure occurred" and "related contract".

[0007] Furthermore, the graph machine learning engine is used to perform risk prediction or intelligent recommendation based on the data in the knowledge graph library; the reinforcement learning engine is used to perform dynamic task scheduling or resource optimization based on the laboratory environment status.

[0008] Furthermore, the graph machine learning engine employs models including graph convolutional networks, graph attention networks, or GraphSAGE; the reward function of the reinforcement learning engine is calculated based on at least one of the following factors: task completion, fault penalty, resource utilization efficiency, and load balancing.

[0009] Furthermore, the AI ​​platform layer also includes a natural language processing engine and an optical character recognition engine; the natural language processing engine is used to parse contract text or generate report content, and the optical character recognition engine is used to extract data from instrument output files.

[0010] Furthermore, the application layer includes a task scheduling module and a detection management module; the task scheduling module calls the reinforcement learning engine to obtain scheduling suggestions, and the detection management module calls the graph machine learning engine to obtain task allocation suggestions.

[0011] According to another aspect of the present invention, an intelligent laboratory management method incorporating graph machine learning is provided; This intelligent lab management approach, which integrates graph machine learning, includes: Construct a laboratory knowledge graph and store laboratory entities and their relationships in a graph structure in the knowledge graph database. The knowledge graph is analyzed using a graph machine learning engine to achieve risk prediction or intelligent recommendation. Reinforcement learning engines are used to dynamically schedule tasks or allocate resources based on the state of the laboratory environment.

[0012] Furthermore, the construction of the laboratory knowledge graph includes: Define node types, including sample, project, inspector, instrument, customer, testing method, and testing task; Define edge types, including at least one of "belongs to", "detects", "assigned to", "uses", "possesses skills", and "related to contracts"; Data is extracted from the business database, process database, and instrument status database and mapped to nodes and edges in the knowledge graph.

[0013] Furthermore, the risk prediction using a graph machine learning engine includes: based on a graph neural network model, aggregating node information through a message passing mechanism, learning node representations, and outputting risk prediction results; when the prediction results exceed a threshold, triggering a risk warning.

[0014] Furthermore, the dynamic task scheduling using a reinforcement learning engine includes: defining a state space containing task queues, instrument status, and personnel load; defining an action space containing task allocation and reallocation; and having a policy network output scheduling actions based on a reward function; wherein the reward function is calculated based on at least one of the following factors: task completion reward, fault penalty, and resource utilization efficiency.

[0015] The beneficial effects of this invention are as follows: By constructing an intelligent platform that integrates graph machine learning and reinforcement learning, artificial intelligence capabilities are deeply embedded into the entire process of laboratory management. This enables the system to achieve seamless integration and intelligent mining of multi-source data, effectively breaking down information silos and significantly improving data consistency and processing efficiency. Furthermore, through dynamic task scheduling and resource optimization, overall operational efficiency and resource utilization are improved. At the same time, by leveraging proactive risk prediction and propagation analysis mechanisms, early identification and intelligent control of potential risks are achieved, enhancing laboratory safety and management scientificity. Ultimately, the goal is to comprehensively improve the level of laboratory automation, intelligence, and optimized management. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a system architecture diagram of an intelligent laboratory management system for fusion graph machine learning according to an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, the intelligent laboratory management system for fusion graph machine learning according to an embodiment of the present invention includes a data layer, an AI middleware layer, an application layer, and a user layer. The data layer is configured to store laboratory management-related data, including a business database, a process database, an instrument status database, and a knowledge graph library. The AI ​​platform layer is connected to the data layer and includes a graph machine learning engine and a reinforcement learning engine. The application layer is connected to the AI ​​platform layer and is used to call the services of the AI ​​platform layer to realize laboratory business functions. The user layer is connected to the application layer and is used to provide interactive interfaces for users with different roles.

[0020] According to an embodiment of the present invention, the intelligent laboratory management system integrating graph machine learning, in a specific implementation, the knowledge graph library is constructed using a graph database, and the stored node types include samples, projects, inspectors, instruments, customers, testing methods, and testing indicators; the stored edge types include at least one of "belongs to", "responsible for testing", "use", "requires qualification", "abnormal result", "fault occurred" and "related contract".

[0021] In a specific embodiment of the intelligent laboratory management system integrating graph machine learning according to an embodiment of the present invention, the graph machine learning engine is used to perform risk prediction or intelligent recommendation based on data in the knowledge graph library; the reinforcement learning engine is used to perform dynamic task scheduling or resource optimization based on the laboratory environment status.

[0022] According to an embodiment of the intelligent laboratory management system integrating graph machine learning, in a specific implementation, the graph machine learning engine uses a model including a graph convolutional network, a graph attention network, or GraphSAGE; the reward function of the reinforcement learning engine is calculated based on at least one of the following factors: task completion status, fault penalty, resource utilization efficiency, and load balancing.

[0023] In a specific embodiment of the intelligent laboratory management system for fusion graph machine learning according to an embodiment of the present invention, the AI ​​platform layer further includes a natural language processing engine and an optical character recognition engine; the natural language processing engine is used to parse contract text or generate report content, and the optical character recognition engine is used to extract data from instrument output files.

[0024] According to an embodiment of the intelligent laboratory management system integrating graph machine learning according to an embodiment of the present invention, in a specific implementation, the application layer includes a task scheduling module and a detection management module; the task scheduling module calls the reinforcement learning engine to obtain scheduling suggestions, and the detection management module calls the graph machine learning engine to obtain task allocation suggestions.

[0025] Secondly, according to an embodiment of the present invention, the intelligent laboratory management method based on fused graph machine learning includes: Construct a laboratory knowledge graph and store laboratory entities and their relationships in a graph structure in the knowledge graph database. The knowledge graph is analyzed using a graph machine learning engine to achieve risk prediction or intelligent recommendation. Reinforcement learning engines are used to dynamically schedule tasks or allocate resources based on the state of the laboratory environment.

[0026] According to an embodiment of the intelligent laboratory management method based on fusion graph machine learning described in this invention, in a specific implementation, the construction of the laboratory knowledge graph includes: Define node types, including sample, project, inspector, instrument, customer, testing method, and testing task; Define edge types, including at least one of "belongs to", "detects", "assigned to", "uses", "possesses skills", and "related to contracts"; Data is extracted from the business database, process database, and instrument status database and mapped to nodes and edges in the knowledge graph.

[0027] According to an embodiment of the present invention, the intelligent laboratory management method integrating graph machine learning, in a specific implementation, includes the following steps for using a graph machine learning engine to achieve risk prediction: based on a graph neural network model, aggregating node information through a message passing mechanism, learning node representations, and outputting risk prediction results; when the prediction results exceed a threshold, triggering a risk warning.

[0028] According to an embodiment of the present invention, the intelligent laboratory management method based on fused graph machine learning, in a specific implementation, the dynamic task scheduling using a reinforcement learning engine includes: defining a state space containing task queues, instrument status, and personnel load; defining an action space containing task allocation and reallocation; and having a policy network output scheduling actions based on a reward function; wherein the reward function is calculated based on at least one of the following factors: task completion reward, fault penalty, and resource utilization efficiency.

[0029] To facilitate understanding of the above technical solutions of the present invention, the following detailed description of the above technical solutions of the present invention is provided through specific embodiments and examples.

[0030] In practical application, the intelligent laboratory management system based on the fusion graph machine learning described in this invention, such as... Figure 1 As shown, the system adopts a layered architecture design, including a data layer, an AI middleware layer, an application layer, and a user layer.

[0031] The data layer, as the foundation of the system, is responsible for the persistent storage and management of various types of laboratory data, providing stable and reliable data services to the upper layers. This layer comprises four core components: Business Database: Stores core business entities and static configuration data of the system. It is implemented using a relational database and supports ACID transaction characteristics. Key data tables include: Quotation Form: (quotation_id, customer_id, project_id, quotation_content, status, creator_id, create_time, update_time), where the quotation_content field stores project and price details in JSON format; Contract Form: (contract_id, quotation_id, customer_id, contract_file, special_requirements, review_status, sign_date, create_time), where the special_requirements field stores the key terms for NLP parsing; Sample table: (sample_id, project_id, sample_name, specification, source, test_items, expiry_date, status, create_time), where the test_items field stores a list of test items in JSON format; Project List: (project_id, project_name, customer_id, priority, deadline, current_progress, status, create_time); Personnel List: (staff_id, name, role, skills, qualifications, current_workload, max_capacity, status), where the skills and qualifications fields store a list of tags in JSON format; Instrumentation: (instrument_id, name, model, purchase_date, status, current_method, maintenance_interval); Process Database: Stores dynamic data and status change records in business processes, enabling full-process traceability. Key data tables include: Task Flow Chart: (task_id, sample_id, instrument_id, assignee_id, task_type, status, created_time, assigned_time, started_time, completed_time); Test Result Table: (result_id, task_id, sample_id, test_item, raw_data_path, calculated_result, unit, reviewer_id, review_time, status, create_time); Review the transaction log: (audit_id, entity_type, entity_id, reviewer_id, review_action, review_comment, review_time); Instrument Status Database: Collects instrument and equipment operating status data via IoT interface. Key data tables include: Real-time status table: (instrument_id, status, current_method, running_hours, last_sample_id, update_time); Historical status log table: (log_id, instrument_id, status, parameters, error_code, log_time), where the parameters field stores detailed parameters in JSON format. Maintenance record sheet: (maintenance_id, instrument_id, maintenance_type, maintenance_date, next_maintenance_date, description, performed_by); Knowledge Graph Repository: Built using a graph database (such as Neo4j), it stores entities and their relationships. Node types include Sample, Project, Staff, Instrument, Customer, TestMethod, and TestItem. Edge relationships include "belongs to," "responsible for testing," "uses," "requires qualification," "abnormal result," "failed," and "related to contract," etc. This repository provides native graph storage processing, flexible schema evolution, built-in graph algorithm support, and high-concurrency read / write capabilities.

[0032] The AI ​​middleware layer, as the intelligent core of the system, provides standardized artificial intelligence services: Graph Machine Learning Engine: Specifically designed for processing graph-structured data in knowledge graphs, employing graph neural network models (including GCN, GAT, GraphSAGE, etc.). This engine uses a message-passing mechanism to aggregate and learn node information, enabling tasks such as risk prediction, intelligent recommendation, and anomaly detection. Specific applications include: Risk propagation prediction: When instrument node status is abnormal, the risk propagation path and impact range are predicted using graph structure. Intelligent recommendation: Based on the "inspector-project-instrument" relationship diagram, it recommends the optimal task allocation scheme. Anomaly detection: Identifying anomaly patterns through node embedding vectors Reinforcement learning engine: Learns the optimal scheduling policy through the interaction between the agent and the environment, including an environment simulator, a policy network, and a reward function calculator. The reward function is defined as: R(s,a,s')=R_completion(s,a,s')+R_penalty(s,a,s')+R_efficiency(s,a,s')+R_balancing(s,a,s')+R_intervention(s,a,s'); Each component corresponds to a task completion reward, a failure penalty, an efficiency reward, a balance reward, and an intervention penalty.

[0033] Natural Language Processing Engine: Enables functions such as contract text parsing and report content generation.

[0034] Optical character recognition engine: Extracts structured data from instrument output files.

[0035] General-purpose algorithm engine: Encapsulates commonly used machine learning algorithms, including classification algorithms (SVM, decision tree, random forest), regression algorithms (linear regression, SVR), clustering algorithms (K-means, DBSCAN), and anomaly detection algorithms (IsolationForest, One-Class SVM).

[0036] The application layer provides specific business function modules: Intelligent workbench: Displays laboratory operation overview and risk warning information; Task scheduling module: invokes the reinforcement learning engine to obtain dynamic scheduling suggestions; Detection Management Module: Calls the graph machine learning engine to obtain task allocation suggestions; Report management module: Utilizes NLP and OCR engines to automatically generate reports; Instrument Management Module: Calls a general algorithm engine for fault prediction; Electronic quotation / contract management module: Uses an NLP engine to achieve intelligent contract parsing.

[0037] The user layer provides access interfaces for users with different roles: Administrator: Has full system privileges, and is responsible for overall overview and risk warnings; Inspector: Uses the inspection management module to receive tasks and enter data; Sampler: Inputs sample information via mobile device; Sales / Account Manager: Use quotation and contract management features; Review experts: Participate in document review online; Clients: View project progress and reports through the portal.

[0038] Core functionality implementation: Dynamic task scheduling based on reinforcement learning: The system abstracts task scheduling into a Markov decision process and trains the agent using a reinforcement learning algorithm. Environment Definition: The state space includes task queue status (task ID, project ID, testing item, priority, estimated duration, deadline, required instrument type, personnel skill requirements, sample expiration date), instrument and equipment status (operating status, load, health, geographical location), personnel status (load, skills, qualifications, efficiency), and project progress and time information. The action space includes task assignment, reassignment, priority adjustment, and waiting operations.

[0039] Reward function design: A multi-objective reward mechanism is adopted. Positive rewards include rewards for timely completion of tasks (weighted according to priority and lead time), rewards for timely delivery of projects, rewards for instrument utilization, and rewards for balanced personnel workload. Negative rewards include penalties for task delays, project delays, resource idleness, human intervention, and fault triggering.

[0040] Algorithm training: Select the PPO or SAC algorithm, conduct millions of interactive training sessions through an environment simulator, pre-train using historical data, and continuously optimize through online learning after deployment.

[0041] Deployment and execution: The trained policy network is deployed as a microservice, responding to state changes in real time to output scheduling decisions, and providing a manual intervention interface through an intelligent workbench.

[0042] Risk warning based on graph machine learning The system uses knowledge graphs and graph neural networks to predict risks. Knowledge graph construction: Nodes include samples, projects, inspectors, instruments, testing tasks, testing methods, and customers; edge relationships include "belongs to," "testing," "assigned to," "used," "possesses skills," "requires testing," and "related to contracts," etc. Data is extracted from various databases using ETL tools and processed using an NLP engine for unstructured text.

[0043] Feature engineering: Define feature vectors for nodes, such as sample nodes (remaining validity days, number of test items, status code, number of historical anomalies), project nodes (days until deadline, priority code, completion percentage, number of associated samples), inspector nodes (current workload, skill matching degree, average task time, number of faults), and instrument nodes (days until next maintenance, failure probability, mean time between failures, parameter anomaly index).

[0044] Risk propagation model: Employing a GNN model, when a node's state is abnormal, risk information propagates along the edges through a message passing mechanism. After multi-layer aggregation, the output risk prediction includes the probability of project delay, the risk of sample expiration, instrument failure time, and the probability of abnormal test results.

[0045] Early warning mechanism: Set business risk thresholds, trigger an early warning when the predicted value exceeds the limit, display the risk propagation path and response suggestions through the intelligent workbench, and continuously collect feedback to optimize the model.

[0046] System interface implementation: The system provides standardized API interfaces: The graph machine learning engine interface (POST / api / ai / graph / risk-prediction) takes the target node and risk type as input and returns the risk score, a list of affected nodes, and response suggestions.

[0047] The reinforcement learning engine interface (POST / api / ai / rl / optimal-scheduling) takes the task to be processed and the available resources as input and returns the optimal scheduling scheme and expected reward.

[0048] NLP & OCR interface (POST / api / ai / nlp / parse-contract): Input contract text, return the parsed key clauses and confidence level.

[0049] The Smart Workbench API (GET / api / dashboard / overview) returns project statistics, risk warnings, and resource utilization data.

[0050] Task scheduling interface (POST / api / scheduling / assign-task): Input task requirements, return allocation results and alternative solutions.

[0051] Report Management Interface (POST / api / reports / generate): Input project information and template, and return automatically generated report content.

[0052] System Deployment and Optimization: The system adopts a microservice architecture, enabling containerized deployment and elastic scaling through Docker and Kubernetes. A model monitoring system tracks performance metrics, and a CI / CD pipeline supports rapid iteration. Continuous model learning and optimization are achieved through user feedback loops.

[0053] In summary, by utilizing the technical solutions described above in this invention, an intelligent platform integrating graph machine learning and reinforcement learning is constructed, deeply embedding artificial intelligence capabilities into the entire laboratory management process. This enables the system to achieve seamless integration and intelligent mining of multi-source data, effectively breaking down information silos and significantly improving data consistency and processing efficiency. Furthermore, through dynamic task scheduling and resource optimization, overall operational efficiency and resource utilization are improved. Simultaneously, by employing proactive risk prediction and propagation analysis mechanisms, early identification and intelligent control of potential risks are achieved, enhancing laboratory safety and management scientificity. Ultimately, the goal is to comprehensively improve the automation, intelligence, and optimized management level of the laboratory.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent laboratory management system integrating graph machine learning, characterized in that, It includes the data layer, AI platform layer, application layer, and user layer; The data layer is configured to store laboratory management-related data, including a business database, a process database, an instrument status database, and a knowledge graph library. The AI ​​platform layer is connected to the data layer and includes a graph machine learning engine and a reinforcement learning engine. The application layer is connected to the AI ​​platform layer and is used to call the services of the AI ​​platform layer to realize laboratory business functions. The user layer is connected to the application layer and is used to provide interactive interfaces for users with different roles.

2. The intelligent laboratory management system for fusion graph machine learning according to claim 1, characterized in that, The knowledge graph database is constructed using a graph database. The types of nodes stored include samples, projects, inspectors, instruments, customers, testing methods, and testing indicators. The types of edges stored include at least one of the following: "belongs to", "responsible for testing", "use", "requires qualification", "abnormal result", "failed", and "related to contract".

3. The intelligent laboratory management system for fusion graph machine learning according to claim 1, characterized in that, The graph machine learning engine is used to perform risk prediction or intelligent recommendation based on the data in the knowledge graph library; the reinforcement learning engine is used to perform dynamic task scheduling or resource optimization based on the laboratory environment conditions.

4. The intelligent laboratory management system for fusion graph machine learning according to claim 3, characterized in that, The graph machine learning engine employs models including graph convolutional networks, graph attention networks, or GraphSAGE; the reward function of the reinforcement learning engine is calculated based on at least one of the following factors: task completion, fault penalty, resource utilization efficiency, and load balancing.

5. The intelligent laboratory management system for fusion graph machine learning according to claim 1, characterized in that, The AI ​​platform layer also includes a natural language processing engine and an optical character recognition engine; the natural language processing engine is used to parse contract text or generate report content, and the optical character recognition engine is used to extract data from instrument output files.

6. The intelligent laboratory management system for fusion graph machine learning according to claim 1, characterized in that, The application layer includes a task scheduling module and a detection management module; the task scheduling module calls the reinforcement learning engine to obtain scheduling suggestions, and the detection management module calls the graph machine learning engine to obtain task allocation suggestions.

7. A smart laboratory management method integrating graph machine learning, applied to the system as described in any one of claims 1 to 6, characterized in that, The method includes: Construct a laboratory knowledge graph and store laboratory entities and their relationships in a graph structure in the knowledge graph database. The knowledge graph is analyzed using a graph machine learning engine to achieve risk prediction or intelligent recommendation. Reinforcement learning engines are used to dynamically schedule tasks or allocate resources based on the state of the laboratory environment.

8. The intelligent laboratory management method based on fused graph machine learning according to claim 7, characterized in that, The construction of the laboratory knowledge graph includes: Define node types, including sample, project, inspector, instrument, customer, testing method, and testing task; Define edge types, including at least one of "belongs to", "detects", "assigned to", "uses", "possesses skills", and "related to contracts"; Data is extracted from the business database, process database, and instrument status database and mapped to nodes and edges in the knowledge graph.

9. The intelligent laboratory management method based on fused graph machine learning according to claim 7, characterized in that, The risk prediction using a graph machine learning engine includes: based on a graph neural network model, aggregating node information through a message passing mechanism, learning node representations, and outputting risk prediction results; when the prediction results exceed a threshold, a risk warning is triggered.

10. The intelligent laboratory management method based on fused graph machine learning according to claim 7, characterized in that, The dynamic task scheduling using a reinforcement learning engine includes: defining a state space containing task queues, instrument status, and personnel load; defining an action space containing task allocation and reallocation; and having a policy network output scheduling actions based on a reward function. The reward function is calculated based on at least one of the following factors: task completion reward, fault penalty, and resource utilization efficiency.

Images