Human-machine-object collaborative scheduling method and system for smart communities based on knowledge graph

Abstract

The embodiment of this specification provides a method and system for the coordinated scheduling of human-machine-object in a smart community based on a knowledge graph, wherein the method includes: collecting data from IoT devices and sensors in the community, transmitting the data to the intelligent system within the community, and establishing a connection between residents and the intelligent system through intelligent terminal devices to create an information and instruction exchange channel; forming a human-machine-object triple structure according to the static relationship between entities, identifying variables with causal relationships based on the collected data of various IoT devices and sensors, constructing a human-machine-object ternary knowledge graph, and maintaining and updating the human-machine-object ternary knowledge graph through a causal reasoning method; based on the human-machine-object ternary knowledge graph, using a deep learning network to select and execute scheduling tasks and feedback the execution results, and updating the knowledge graph according to the execution results. The present invention can improve the efficiency of scheduling task processing and enhance the overall intelligence level of the community.

Description

Technical Field

[0001] This document relates to the field of smart community technology, and in particular to a method and system for smart community human-machine-object collaborative scheduling based on knowledge graph. Background Art

[0002] The fusion of man, machine and object represents the deep integration of physical space, information space and social space, in which the physical space constantly interacts closely with the information space and social space, while the information space and social space achieve a high degree of integration of cognitive attributes and computational attributes. In smart communities, property services are the key link in this fusion. By combining people, machines and IoT devices, intelligent management and efficient scheduling of property services can be achieved. For example, sensors are used to monitor the community environment in real time, and resource allocation is optimized through data analysis to improve service response speed and ensure the safety and comfort of community residents. However, the existing smart community dispatching methods still face several challenges and shortcomings: (1) The current smart community dispatching system is often unable to respond immediately when dealing with emergencies and lacks a fast and efficient dispatching mechanism, resulting in the urgent needs of community residents not being met in a timely manner; (2) The information between various systems and devices is still in a decentralized state. The lack of an effective data integration mechanism makes it impossible to share and utilize important information, thereby reducing dispatching efficiency and making it difficult to meet the unified and efficient joint mobilization of the central control system; (3) Most existing dispatching systems rely on preset rules to make decisions, which leads to insufficient adaptability and flexibility of the system when faced with unknown or complex scenarios, making it difficult to quickly adjust and respond to personalized needs and unable to provide considerate services; (4) In the operation of smart communities, the collection, storage and processing of large amounts of sensitive data brings significant security and privacy risks, increasing the possibility of data leakage and abuse. Summary of the invention

[0003] One or more embodiments of this specification provide a method for human-machine-object collaborative scheduling in a smart community based on a knowledge graph, including:

[0004] Deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the community's internal intelligent system. Residents establish connections with the community's internal intelligent system through intelligent terminal devices to create information and command exchange channels;

[0005] According to the static relationship between entities, a human-machine-object triple structure is formed. Based on the data collected from various IoT devices and sensors, the variables with causal relationships in the triple structure are identified, and a human-machine-object ternary knowledge graph with entity-relationship attributes is constructed. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods.

[0006] Based on the human-machine-object ternary knowledge graph, a deep learning network is used to select execution scheduling tasks and feedback the execution results, and the knowledge graph is updated according to the execution results.

[0007] Furthermore, the sensors include temperature and humidity sensors, motion sensors and light sensors, which are used to monitor the physical environment of the community and record time nodes through time and date monitoring.

[0008] Furthermore, residents establish connections with the community's internal intelligent systems through intelligent terminal devices, creating information and command exchange channels, including:

[0009] Residents use smart terminal devices to establish connections with the community's smart system, and the smart terminal devices include smart phones, tablets and computers;

[0010] Sending specific instruction information or request information to the intelligent system through the intelligent terminal device;

[0011] The intelligent system processes the received instruction information or demand information according to preset algorithms and conditions, converts it into corresponding control signals, and sends it to the corresponding IoT device to perform corresponding operations.

[0012] Furthermore, the sending of specific instruction information or raising of demand information to the intelligent system through the intelligent terminal device specifically includes:

[0013] Send instructions or demand information to the intelligent system through the graphical user interface GUI or voice assistant of the intelligent terminal device.

[0014] The intelligent system converts the received instructions or demand information into word vectors, matches similar instructions or demands in the human-machine-object ternary knowledge graph, and connects to the equipment and corresponding parameters associated with the demand. After obtaining the equipment and parameters of the demand being processed, the user input is converted into clear control instructions.

[0015] Furthermore, a human-machine-object ternary knowledge graph with entity-relationship attributes is constructed, and the constructed human-machine-object ternary knowledge graph is maintained and updated through the causal reasoning method. The specific method is:

[0016] A probabilistic graph model is constructed through a constrained Bayesian network to represent the causal dependency between variables. In the probabilistic graph model, nodes represent entities or events in the community, and directed edges represent the potential impact of one entity on another entity. The Bayesian network automatically identifies conditional independence through an algorithm to obtain a true causal structure.

[0017] Set constraints on the Bayesian network to limit direct connections between variables that do not have a causal relationship;

[0018] The result of the conditional independence test is obtained by automatically testing the conditional independence between variables, including: based on the Bayesian rule, judging whether two variables are independent given other variables, and if so, removing the connection between the two nodes;

[0019] Generate a directed acyclic graph (DAG) based on the results of the conditional independence test, where each directed edge represents the causal dependency between two variables. By maximizing the posterior probability, the optimal causal structure is determined.

[0020] Using maximum likelihood estimation or Bayesian estimation, parameters are estimated for each edge in the directed acyclic graph DAG to quantify the strength of the causal relationship.

[0021] Furthermore, the method further comprises:

[0022] The causal forest algorithm is applied to estimate the individual treatment effect ITE. The causal forest constructs multiple decision trees to evaluate the expected results under specific conditions for each individual and perform personalized causal effect estimation.

[0023] Furthermore, the construction and application process of the causal forest algorithm is as follows:

[0024] Data preparation and preprocessing: The system collects observation data variables of different residents or IoT devices in the community, standardizes the collected observation data variables and handles missing values. The observation data variables include intervention measures and corresponding results;

[0025] Construct random decision tree clusters: Use different subsets of the data to construct multiple decision trees by random sampling. Each decision tree estimates the causal effect of the intervention and generates an individual treatment effect ITE associated with a specific device or resident.

[0026] Individual effect estimation: For each individual, including IoT devices or residents, the causal forest algorithm is used to estimate the expected treatment effects under different conditions and identify the personalized impact of specific interventions on a device or resident.

[0027] Integrate the results of multiple decision trees: By integrating the estimated results of all decision trees, the system takes a weighted average of the treatment effects of each individual to obtain the integrated results;

[0028] Causal effect analysis and explanation: Through the integrated results, the intelligent system automatically generates the distribution of individual treatment effects ITE and provides personalized causal explanations based on the characteristics of IoT devices or residents.

[0029] One or more embodiments of this specification provide a knowledge graph-based smart community human-machine-object collaborative scheduling system, including:

[0030] Information collection module: used to deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the intelligent system within the community. Residents establish connections with the intelligent system within the community through intelligent terminal devices to create information and command exchange channels;

[0031] Knowledge graph construction module: used to form a human-machine-object triple structure based on the static relationship between entities, identify variables with causal relationships in the triple structure based on the data collected from various IoT devices and sensors, and construct a human-machine-object ternary knowledge graph with entity-relationship attributes. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods;

[0032] Collaborative scheduling module: It is used to select execution scheduling tasks and feedback the execution results based on the human-machine-object ternary knowledge graph, and update the knowledge graph according to the execution results.

[0033] One or more embodiments of the present specification provide an electronic device, including:

[0034] processor; and,

[0035] A memory arranged to store computer-executable instructions, which, when executed, enable the processor to implement the steps of the above-mentioned knowledge graph-based smart community human-machine-object collaborative scheduling method.

[0036] One or more embodiments of the present specification provide a storage medium for storing computer-executable instructions, which, when executed, implement the steps of the above-mentioned knowledge graph-based smart community human-machine-object collaborative scheduling method.

[0037] By adopting the embodiment of the present invention, a knowledge network is formed by constructing a knowledge graph of property services in a smart community, which can more accurately match tasks and entities, match events through preset scenario classifications, and accurately judge and understand the current environment and context; the knowledge graph is used to deeply analyze the relationship between events and other entities and the interaction in different scenarios, and the relationship between users' potential needs and actual needs is explored, so that community services can be efficiently scheduled in advance or immediately to meet users' potential needs, thereby improving the efficiency of task and event processing and enhancing the overall intelligence level.

[0038] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 A flowchart of a method for human-machine-object collaborative scheduling in a smart community based on a knowledge graph provided for one or more embodiments of this specification;

[0041] Figure 2 A schematic diagram of a human-machine-object interaction process framework provided for one or more embodiments of this specification;

[0042] Figure 3 A schematic diagram of the structure of a community collaborative intelligent scheduling method integrating a knowledge graph provided in one or more embodiments of this specification;

[0043] Figure 4 A schematic diagram of the composition of a knowledge graph-based smart community human-machine-object collaborative scheduling system provided for one or more embodiments of this specification;

[0044] Figure 5 A schematic diagram of the structure of an electronic device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0046] Method Embodiment

[0047] According to an embodiment of the present invention, a method for human-machine-object collaborative scheduling in a smart community based on a knowledge graph is provided. Figure 1 A flowchart of a method for collaborative scheduling of human-machine-object in a smart community based on a knowledge graph is provided for one or more embodiments of this specification, such as Figure 1 As shown, the method for human-machine-object collaborative scheduling in a smart community based on a knowledge graph according to an embodiment of the present invention specifically includes:

[0048] S1. Deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the intelligent system within the community. Residents establish connections with the intelligent system within the community through smart terminal devices to create information and command exchange channels.

[0049] In the process of human-machine-object interaction in smart communities, the interaction between residents and intelligent systems is multi-level and multi-dimensional. In the smart community scenario, residents establish connections with the intelligent systems within the community through intelligent terminal devices, such as smartphones or tablets, and create a channel through this connection to exchange information and instructions between the two parties. When residents issue instructions or make demands through intelligent terminal devices that have established a connection with the intelligent system, these instructions or demands are transmitted to the intelligent system through the network and converted into recognizable control signals after processing. They are then sent to the corresponding objects according to preset algorithms and conditions to control the objects. After receiving the control signal, the object responds accordingly to meet the needs of the residents. The human-machine-object interaction process is as follows: Figure 2 As shown, the details are as follows:

[0050] Residents use smart terminal devices such as smartphones or tablets to establish connections with the community's smart systems. This connection is usually achieved through wireless networks, including Wi-Fi or other mobile communication technologies. When residents' smart terminal devices establish connections with the smart systems and log in, they go through a secure authentication and authorization process to ensure data security and privacy protection.

[0051] After the connection is established, residents can send request information to the intelligent system through a graphical user interface (GUI) or voice assistant, including specific instructions or demands. These request information are sent to the central processing unit of the intelligent system through the network. The central processing unit processes the demands as follows: the intelligent system converts the received instructions or demand information into word vectors, matches similar instructions or demands in the human-machine-object ternary knowledge graph, and connects to the equipment and corresponding parameters associated with the demand. After obtaining the equipment and parameters of the demand being processed, the user input is converted into clear control instructions.

[0052] After receiving the control instructions, the intelligent system processes the data according to the preset algorithms and conditions, where the preset algorithms include machine learning models and complex decision-making logic, and determines how to respond to residents' needs through the preset algorithms. For example, after the preset algorithms and conditions are processed, the residents' needs may be to adjust the temperature, turn on and off the lights, or start the security monitoring system.

[0053] The processed instructions are converted into control signals and sent to related IoT devices through the network, allowing these devices to perform corresponding operations, such as adjusting the air-conditioning temperature, controlling door switches, or operating surveillance cameras.

[0054] During the interaction process, the intelligent system continuously monitors the physical environment through various sensors, such as temperature and humidity sensors, motion sensors, and light sensors, as well as paying attention to time nodes through time and date monitoring. Real-time data is obtained through these sensors to help the system understand the living environment and immediate needs of residents more accurately. At the same time, these sensors themselves also have scheduling parameter attributes. Special situations in the monitoring will trigger the scheduling mechanism, such as adopting different scheduling plans according to different holidays, or using physical sensors to dispatch multiple devices to a suitable temperature for residents when parameters such as corridor temperature change.

[0055] According to sensor data and residents' interactive feedback, the intelligent system can automatically adjust its response strategy. Specifically, the intelligent system obtains users' evaluations of each scheduling decision at a fixed time, weighs the complexity of residents' lives and the mild disturbance of the survey, and sets the fixed time to obtain users' evaluations of each scheduling decision in the form of a questionnaire once a week. The intelligent system analyzes the residents' usage habits and feedback in the questionnaire, and through the reinforcement learning module in the system, punishes the scheduling with poor feedback and rewards the scheduling with good feedback, continuously learning and optimizing, and improving service quality and residents' satisfaction.

[0056] During the entire human-machine-object interaction process, the intelligent system uses encryption technology to protect data transmission, and all user data processing complies with privacy protection regulations and standards to ensure the security of data transmission.

[0057] S2. According to the static relationship between entities, a human-machine-object triple structure is formed. Based on the data collected from various IoT devices and sensors, the variables with causal relationships in the triple structure are identified, and a human-machine-object ternary knowledge graph with entity-relationship attributes is constructed. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods.

[0058] When constructing a human-machine-object ternary knowledge graph through causal reasoning methods, first identify and select variables with potential causal relationships from the existing data. The purpose of variable identification is to determine which data attributes may have explanatory power for the phenomenon being studied, or there may be an influence relationship between people, machines, and objects in the community. In this embodiment, by combining domain knowledge and data exploration technology, it is predicted which variables may have a causal relationship, specifically including:

[0059] Application of domain knowledge. Through cooperation with community managers, residents, equipment suppliers, etc., we collect their understanding of community operations and their intuitive views on which factors may affect each other, and consult relevant research or case studies. Among them, we can query the scheduling logs of community-related management, etc., to understand which variables in similar communities have been proven to have important interactions. This step is mainly information mining. The mined information, such as the correlation between the temperature x1 of air conditioner 1 on the 1F floor and the temperature x2 of air conditioner 2, etc., on the temperature comfort of the corridor on this floor. The background of this example is that the entry and exit time of the residents in the corridor area of ​​this floor and the main areas they pass through are basically fixed. The control of different air conditioning time and temperature can actually accurately control the appropriate temperature of each area, which can not only save energy, but also control the temperature of each household door area according to the user's temperature requirements. The core is to use intelligent systems to push scheduling to be as humane as possible for each user.

[0060] Data exploration. Use statistical methods to identify variables with causal relationships based on the various IoT devices and sensors collected. This embodiment calculates the Pearson correlation coefficient between variables to preliminarily analyze the correlation between the variables, helping to narrow the scope of potential causal variables.

[0061] When analyzing the correlation between community IoT device parameters and residents' needs, the mutual information method is used to evaluate the dependence between device parameters. It is defined as the amount of common information between two random variables X and Y, and its calculation formula is as follows:

[0062] ;

[0063] in is the joint probability distribution of variables X and Y; and are their respective marginal probability distributions; the higher the common information (mutual information value), the stronger the correlation; by calculating the mutual information value between equipment parameters, the variables with the most significant correlation with household needs are screened out.

[0064] The mutual information method can not only measure linear relationships, but also capture nonlinear correlations, thereby providing a more comprehensive understanding of the mutual correlations between device parameters.

[0065] By using the causal discovery algorithm (PC algorithm), the causal relationship in the knowledge graph of smart community property services is deeply analyzed and verified. The PC algorithm is an automated causal discovery tool that can identify and model potential causal connections from complex community data. The algorithm gradually constructs a causal network between device parameters through conditional independence tests, showing the interaction relationship between explicit and latent variables.

[0066] First, the present invention inputs the collected equipment parameter data set based on the previous data preprocessing and variable identification results, combined with the insights of field experts and relevant literature research. The PC algorithm automatically generates a cause-effect relationship diagram through the following process:

[0067] Initialize the fully connected graph: Initially, assume that there is a causal relationship between all variables (device parameters) and construct a fully connected undirected graph;

[0068] Independence test: Step by step, check whether each pair of variables is conditionally independent. If two variables X and Y are independent under the condition of a given variable set Z, remove the edge between X and Y. The test is based on the conditional independence test. The formula is as follows:

[0069] P(X,Y∣Z)=P(X∣Z)P(Y∣Z);

[0070] If this independence condition is met, then X and Y have no direct causal relationship;

[0071] Determine directionality: Based on the independence test results, use the directional rule to direct the undirected edges and construct a directed acyclic graph (DAG) to represent the causal structure between device parameters;

[0072] After the causal graph is constructed, the present invention compares the generated causal graph with the domain expert hypothesis and verifies the rationality of the model. The causal graph shows the directed edges between the device parameter nodes, indicating the direction of the causal relationship and the strength of the interaction, helping to determine which device parameters have a significant impact on the needs of residents.

[0073] Construct a human-machine-object ternary knowledge graph with entity-relationship attributes, and maintain and update the constructed human-machine-object ternary knowledge graph through causal reasoning method. The specific method is:

[0074] Use causal reasoning algorithms to build a human-machine-object knowledge graph for smart communities. Specifically:

[0075] A constrained Bayesian network is selected to construct a probabilistic graph model to represent the causal dependency between variables. In the probabilistic graph model, nodes represent entities or events in the community, and directed edges represent the potential impact of one entity on another. The Bayesian network automatically identifies conditional independence through an algorithm to obtain a true causal structure; this process does not rely on any prior assumptions, which improves the objectivity and accuracy of the model. Definition of nodes and variables: First, define the nodes in the network according to the entities involved in the community (such as equipment, sensors, and resident needs). Each node corresponds to a specific variable or event. For example: Node: 1F corridor resident temperature demand, 101 indicator xx... Connection variables: air conditioner 1 temperature x1; air conditioner 2 temperature x2... indoor temperature sensor temperature T1;

[0076] Set the constraints of the Bayesian network to limit the direct connection between variables that do not have a causal relationship, so as to ensure a more reasonable network structure. For example, there should be no direct causal relationship between certain devices, such as indoor corridor temperature and community street lights.

[0077] The result of the conditional independence test is obtained by automatically testing the conditional independence between variables, including: based on the Bayesian rule, judging whether two variables are independent given other variables, and if so, removing the connection between the two nodes;

[0078] Generate a directed acyclic graph (DAG) based on the results of the conditional independence test, where each directed edge represents the causal dependency between two variables. By maximizing the posterior probability, the optimal causal structure is determined.

[0079] Using maximum likelihood estimation or Bayesian estimation, parameters are estimated for each edge in the directed acyclic graph DAG to quantify the strength of the causal relationship.

[0080] Secondly, the present invention applies a causal forest algorithm to estimate the individual treatment effect (ITE). The causal forest constructs multiple decision trees to evaluate the expected results under specific conditions for each individual, thereby providing a more personalized causal effect estimate.

[0081] In actual operation, the construction and application process of the causal forest algorithm is as follows:

[0082] Data preparation and preprocessing: First, the system collects observation data of different residents or IoT devices in the community, including intervention measures (such as device status adjustment) and corresponding results (such as changes in resident satisfaction). In the data preprocessing stage, the variables are standardized and missing values ​​are processed to ensure that the model can obtain reliable causal effect estimates. The observation data variables include intervention measures and corresponding results.

[0083] Construct clusters of randomized decision trees: Multiple decision trees are constructed using different subsets of the data (by random sampling). Each decision tree estimates the causal effect of the intervention and generates an individual treatment effect (ITE) associated with a specific facility or resident.

[0084] Individual effect estimation: For each individual (IoT device or resident), the causal forest algorithm is used to estimate the expected treatment effect under different conditions on multiple trees. By analyzing these individual treatment effects, the personalized impact of a specific intervention on a device or resident can be identified.

[0085] Integrate the results of multiple decision trees: By integrating the estimated results of all decision trees, the system takes a weighted average of the treatment effects of each individual to obtain the integrated results. This process can reduce the bias of single decision tree estimation and provide more stable and accurate causal effect estimates.

[0086] Causal effect analysis and explanation: Based on the integrated results, the intelligent system automatically generates the distribution of individual treatment effects (ITE) and provides personalized causal explanations based on the characteristics of the equipment or residents. This analysis can help understand the differences in the effects of specific interventions under different conditions and provide a reference for optimizing the operation of community equipment.

[0087] Through the causal reasoning method, a knowledge graph containing rich causal relationships is constructed. This graph can not only accurately reflect the various dynamic relationships within the community, but also adapt to the needs of community development and change by dynamically updating these relationships. Ultimately, the knowledge graph based on causal reasoning can mine the correlation patterns between various devices and parameters in community management and user needs or potential needs, making the community management system more humane and able to meet potential needs or efficiently meet user needs through pre-regulation.

[0088] S3. Based on the human-machine-object ternary knowledge graph, a deep learning network is used to select execution scheduling tasks and feedback the execution results, and the knowledge graph is updated according to the execution results.

[0089] This embodiment takes the human-machine-object ternary knowledge graph as the research object to realize the intelligent scheduling management of human-machine-object interaction. By analyzing the information stored in the knowledge graph and using advanced reasoning algorithms, appropriate intelligent strategies are formulated for scheduling tasks in different scenarios. Specifically, for scenarios with discrete action spaces, the present invention uses the Deep Q-Network (DQN) technology for processing; and for scenarios that require continuous scheduling decisions, the Deep Deterministic Policy Gradient (DDPG) method is applied to solve them.

[0090] In DQN and DDPG, although both are based on deep neural networks to learn control strategies, the network structures are different. DQN uses a separate Q network to input the current state (such as corridor temperature, outside temperature, resident activities, etc.), and outputs the Q value of each discrete action (such as "increase temperature", "lower temperature"), representing the expected value of these actions. By selecting the action with the highest Q value, DQN realizes the decision of temperature control. DDPG adopts the actor-critic architecture, in which the policy network (Actor) outputs a continuous action value (such as the specific amplitude of temperature adjustment) according to the current state, and the value network (Critic) evaluates the long-term value of the action selected by the policy network to help guide the optimization of the policy network. Both structures continuously update the network parameters through back propagation in deep learning, so that DQN and DDPG can make the best decision in different temperature control scenarios.

[0091] DQN technology combines the advantages of deep learning and reinforcement learning. By adopting the optimal behavior value function of reinforcement learning and adding an experience replay mechanism, it effectively solves the problem of sample correlation during training. The key to this mechanism is to accumulate historical experience in a buffer and randomly extract samples from it for training. This can destroy the time series association between samples and enhance the diversity of data. DQN is suitable for decision-making problems with discrete action spaces, especially in situations where the best action needs to be selected from many potential options. For example, in smart community property services, DQN can be used to execute a series of scheduling decisions for preset actions, such as controlling the opening and closing of community lighting equipment, managing the usage time of public facilities, and scheduling cleaning and maintenance tasks. By analyzing historical data, DQN can learn and optimize these decision-making processes to improve the efficiency and response speed of property services. The DQN control scheduling process is as follows:

[0092] The process of the DQN system controlling the corridor temperature system starts with state definition. The system determines the state of the environment based on sensor information such as the current corridor temperature, time, outside temperature, and resident activity. These states help DQN understand the real-time status of the corridor, such as whether the corridor temperature is appropriate, whether the resident is passing through the corridor, etc., so as to select the most appropriate action for each state.

[0093] In specific operations, the DQN system adjusts the corridor temperature through a discrete action space. The optional actions include "increase temperature", "lower temperature" or "maintain current temperature", which are automatically generated according to the current state. During periods of high occupant activity, the DQN system can adjust the temperature to ensure comfort, and when no one is active, it will lower the temperature to save energy.

[0094] To optimize the decision-making process, the DQN system sets a reward function to measure the quality of each action. For example, when the corridor maintains a suitable temperature when residents are active, the system receives a positive reward; if too much energy is consumed when no one is active, it will bring a negative reward. Through this reward mechanism, the DQN system continuously adjusts its strategy and learns how to balance the comfort of residents and energy efficiency in different environments.

[0095] The DQN system further enhances the system's learning ability through the experience replay mechanism. Each time the corridor temperature is adjusted, the system stores the current state, action, and reward in a buffer, and randomly extracts these historical experiences for learning in subsequent training. This mechanism can break the continuity in time, increase the diversity of data, and help the system better adapt to different corridor states.

[0096] With multiple trainings and accumulated experience, the DQN system gradually optimizes its strategy and learns how to choose the best temperature adjustment action under different conditions. Finally, after sufficient training, the DQN system can automatically select the appropriate temperature adjustment strategy according to the current corridor status in a real-time environment. For example, when residents pass through the corridor, the system will automatically increase the temperature to ensure comfort; at night or when no one is active, the system will choose to lower the temperature or turn off the heating equipment to maximize energy conservation.

[0097] The DDPG (Deep Deterministic Policy Gradient) algorithm is based on DQN technology, retains the characteristics of experience replay, stores past experience in a buffer, and randomly extracts samples from the buffer for training to reduce the correlation between samples and improve learning efficiency. DDPG combines the policy gradient method and introduces a policy network (Actor) to directly output the numerical value of the action, which is in contrast to the method of DQN outputting the probability distribution of the action, making DDPG particularly suitable for problems in continuous action space. The difference from DQN is that DQN can only learn discrete, that is, fixed operations, while DDPG can learn operations within a certain range. In smart community property services, DDPG can be used in scenarios that require continuous scheduling decisions, such as temperature control of intelligent thermostatic systems, environmental monitoring and regulation of community public areas, and real-time path optimization of property service personnel. By combining real-time data of environmental status, DDPG can directly output the optimal scheduling strategy, thereby achieving more accurate and efficient services.

[0098] DDPG adopts a dual network structure, one for the value network (Critic) and the other for the policy network (Actor), forming an Actor-Critic framework. This framework allows DDPG to learn deterministic strategies in a continuous action space and map the environment state to the action, greatly expanding the scope of application of DQN and being suitable for continuous scheduling decision-making scenarios, such as intelligent lighting adjustment in property services, real-time monitoring and adjustment of community security systems, and maintenance and maintenance of property equipment. In addition, the DDPG algorithm sets a slow-updated target network for each of the value network and the policy network to stabilize the target value during the learning process, reduce the fluctuation of the target value, and enhance the stability and convergence speed of the algorithm.

[0099] Through these features, DDPG provides a powerful tool for dealing with complex problems in continuous action space, and can achieve efficient and accurate decision-making in various continuous control tasks. Its application in fields such as automated control, robotics, and intelligent system optimization further proves its practical value as an efficient continuous decision-making algorithm.

[0100] The human-machine-object ternary coordinated scheduling method of the present invention is as follows Figure 3 As shown in the figure, we first use the human-machine-object ternary knowledge graph to improve the intelligence and automation level of the entire system. This knowledge graph forms a comprehensive and dynamic information network by integrating detailed information and data of the three major entity categories of human, machine, and object. This network not only stores the static relationships between entities, such as location and functional attributes, but also captures their dynamic interactions, such as behavior patterns and state changes.

[0101] Through the application of reinforcement learning technology, the human-machine-object ternary knowledge graph can update and optimize the decision-making process in real time. For example, when faced with specific task assignments, the system can automatically analyze the current environmental conditions and residents' needs, and understand the dependencies and interactions between different entities (people, machines, objects) through the knowledge graph. This deep understanding enables the system to predict possible problems and needs in various scenarios, thereby accurately matching tasks and entities to ensure the efficiency and accuracy of task execution. Furthermore, through the event scene judger, real-time data is matched with predefined scene classifications (such as daily life, emergency events, equipment maintenance, etc.) to judge the current environment and context. Such intelligent classification helps the system better understand the nature and urgency of the event, ensure the timeliness and relevance of the response, and also fine-tune the decision-making to the actual needs of users in real time, so that smart property services can be optimally adapted to users as much as possible.

[0102] The beneficial effects of the present invention are as follows:

[0103] The present invention constructs a knowledge graph of property services in a smart community to form a knowledge network, which can more accurately match tasks and entities, match events through preset scenario classifications, and accurately judge and understand the current environment and context. It uses the knowledge graph to deeply analyze the relationship between events and other entities and the interactions in different scenarios, and explores the relationship between users' potential needs and actual needs, so that community services can be efficiently scheduled in advance or in real time to meet users' potential needs, thereby improving the efficiency of task and event processing and enhancing the overall intelligence level.

[0104] System Example

[0105] According to an embodiment of the present invention, a knowledge graph-based smart community human-machine-object collaborative scheduling system is provided. Figure 4 A schematic diagram of the composition of a knowledge graph-based smart community human-machine-object collaborative scheduling system provided for one or more embodiments of this specification, such as Figure 4 As shown, the human-machine-object collaborative scheduling system for a smart community based on a knowledge graph according to an embodiment of the present invention specifically includes:

[0106] Information collection module 40: used to deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit the data to the intelligent system within the community. Residents establish a connection with the intelligent system within the community through intelligent terminal devices to create an information and command exchange channel;

[0107] Knowledge graph construction module 42: used to form a human-machine-object triple structure according to the static relationship between entities, identify variables with causal relationships based on various types of IoT devices and sensors collected, construct a human-machine-object ternary knowledge graph with entity-relationship attributes, and maintain and update the constructed human-machine-object ternary knowledge graph through causal reasoning methods;

[0108] Collaborative scheduling module 44: based on the human-machine-object ternary knowledge graph, a deep learning network is used to select execution scheduling tasks and feedback the execution results, and the knowledge graph is updated according to the execution results.

[0109] The embodiment of the present invention is a system embodiment corresponding to the above-mentioned method embodiment. The specific operations of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0110] Device Example 1

[0111] An embodiment of the present invention provides an electronic device, such as Figure 5 As shown, it includes: a memory 50, a processor 52, and a computer program stored in the memory 50 and executable on the processor 52. When the computer program is executed by the processor 52, the following method steps are implemented:

[0112] S1. Deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the community's internal intelligent system. Residents establish connections with the community's internal intelligent system through intelligent terminal devices to create information and command exchange channels;

[0113] S2. Based on the static relationship between entities, a human-machine-object triple structure is formed. Based on the collected IoT devices and sensors, variables with causal relationships are identified to construct a human-machine-object ternary knowledge graph with entity-relationship attributes. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods.

[0114] S3. Based on the human-machine-object ternary knowledge graph, a deep learning network is used to select execution scheduling tasks and feedback the execution results, and the knowledge graph is updated according to the execution results.

[0115] Device Example 2

[0116] An embodiment of the present invention provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by the processor 52, the following method steps are implemented:

[0117] S1. Deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the community's internal intelligent system. Residents establish connections with the community's internal intelligent system through intelligent terminal devices to create information and command exchange channels;

[0118] S2. Based on the static relationship between entities, a human-machine-object triple structure is formed. Based on the collected IoT devices and sensors, variables with causal relationships are identified to construct a human-machine-object ternary knowledge graph with entity-relationship attributes. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods.

[0119] S3. Based on the human-machine-object ternary knowledge graph, a deep learning network is used to select execution scheduling tasks and feedback the execution results, and the knowledge graph is updated according to the execution results.

[0120] The computer-readable storage medium in this embodiment includes, but is not limited to, ROM, RAM, magnetic disk or optical disk, etc.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for human-machine-object collaborative scheduling in a smart community based on knowledge graph, characterized in that: include: Deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the community's internal intelligent system. Residents establish connections with the community's internal intelligent system through intelligent terminal devices to create information and command exchange channels; According to the static relationship between entities, a human-machine-object triple structure is formed. Based on the data collected from various IoT devices and sensors, the variables with causal relationships in the triple structure are identified, and a human-machine-object ternary knowledge graph with entity-relationship attributes is constructed. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods. Among them, a human-machine-object ternary knowledge graph with entity-relationship attributes is constructed, and the constructed human-machine-object ternary knowledge graph is maintained and updated through the causal reasoning method. The specific method is: A probabilistic graph model is constructed through a constrained Bayesian network to represent the causal dependency between variables. In the probabilistic graph model, nodes represent entities or events in the community, and directed edges represent the potential impact of one entity on another entity. The Bayesian network automatically identifies conditional independence through an algorithm to obtain a true causal structure. Set constraints on the Bayesian network to limit direct connections between variables that do not have a causal relationship; The result of the conditional independence test is obtained by automatically testing the conditional independence between variables, including: based on the Bayesian rule, judging whether two variables are independent given other variables, and if so, removing the connection between the two nodes; Generate a directed acyclic graph (DAG) based on the results of the conditional independence test, where each directed edge represents the causal dependency between two variables. By maximizing the posterior probability, the optimal causal structure is determined. Using maximum likelihood estimation or Bayesian estimation, the parameters of each edge in the directed acyclic graph DAG are estimated to quantify the strength of the causal relationship. Based on the human-machine-object ternary knowledge graph, a deep learning network is used to select execution scheduling tasks and feedback the execution results, and the knowledge graph is updated according to the execution results.

2. The method according to claim 1, characterized in that The sensors include temperature and humidity sensors, motion sensors and light sensors, which are used to monitor the physical environment of the community and record time nodes through time and date monitoring.

3. The method according to claim 2, characterized in that Residents use smart terminal devices to connect with the community's internal smart systems and create information and command exchange channels, including: Residents use smart terminal devices to establish connections with the community's smart system, and the smart terminal devices include smart phones, tablets and computers; Sending specific instruction information or request information to the intelligent system through the intelligent terminal device; The intelligent system processes the received instruction information or demand information according to preset algorithms and conditions, converts it into corresponding control signals, and sends it to the corresponding IoT device to perform corresponding operations.

4. The method according to claim 3, characterized in that The sending of specific instruction information or request information to the intelligent system through the intelligent terminal device specifically includes: Send instructions or demand information to the intelligent system through the graphical user interface GUI or voice assistant of the intelligent terminal device; The intelligent system converts the received instructions or demand information into word vectors, matches similar instructions or demands in the human-machine-object ternary knowledge graph, and connects to the IoT devices and corresponding parameters associated with the demand. After obtaining the IoT devices and parameters that process the demand, the user input is converted into clear control instructions.

5. The method according to claim 1, characterized in that The method further comprises: The causal forest algorithm is applied to estimate the individual treatment effect ITE. The causal forest constructs multiple decision trees to evaluate the expected results under specific conditions for each individual and perform personalized causal effect estimation.

6. The method according to claim 5, characterized in that The construction and application process of the causal forest algorithm is as follows: Data preparation and preprocessing: The system collects observation data variables of different residents or IoT devices in the community, standardizes the collected observation data variables and handles missing values. The observation data variables include intervention measures and corresponding results; Construct random decision tree clusters: Use different subsets of the data to construct multiple decision trees by random sampling. Each decision tree estimates the causal effect of the intervention and generates an individual treatment effect ITE associated with a specific device or resident. Individual effect estimation: For each individual, including IoT devices or residents, the causal forest algorithm is used to estimate the expected treatment effects under different conditions and identify the personalized impact of specific interventions on a device or resident. Integrate the results of multiple decision trees: By integrating the estimated results of all decision trees, the system takes a weighted average of the treatment effects of each individual to obtain the integrated results; Causal effect analysis and explanation: Through the integrated results, the intelligent system automatically generates the distribution of individual treatment effects ITE and provides personalized causal explanations based on the characteristics of IoT devices or residents.

7. A knowledge graph-based smart community human-machine-object collaborative scheduling system, characterized in that: include: Information collection module: used to deploy IoT devices and sensors in the community, collect data from various IoT devices and sensors in the community, and transmit them to the intelligent system within the community. Residents establish connections with the intelligent system within the community through intelligent terminal devices to create information and command exchange channels; Knowledge graph construction module: used to form a human-machine-object triple structure based on the static relationship between entities, identify variables with causal relationships in the triple structure based on the data collected from various IoT devices and sensors, and construct a human-machine-object ternary knowledge graph with entity-relationship attributes. The constructed human-machine-object ternary knowledge graph is maintained and updated through causal reasoning methods; Among them, a human-machine-object ternary knowledge graph with entity-relationship attributes is constructed, and the constructed human-machine-object ternary knowledge graph is maintained and updated through the causal reasoning method. The specific method is: A probabilistic graph model is constructed through a constrained Bayesian network to represent the causal dependency between variables. In the probabilistic graph model, nodes represent entities or events in the community, and directed edges represent the potential impact of one entity on another entity. The Bayesian network automatically identifies conditional independence through an algorithm to obtain a true causal structure. Set constraints on the Bayesian network to limit direct connections between variables that do not have a causal relationship; The result of the conditional independence test is obtained by automatically testing the conditional independence between variables, including: based on the Bayesian rule, judging whether two variables are independent given other variables, and if so, removing the connection between the two nodes; Generate a directed acyclic graph (DAG) based on the results of the conditional independence test, where each directed edge represents the causal dependency between two variables. By maximizing the posterior probability, the optimal causal structure is determined. Using maximum likelihood estimation or Bayesian estimation, the parameters of each edge in the directed acyclic graph DAG are estimated to quantify the strength of the causal relationship. Collaborative scheduling module: It is used to select execution scheduling tasks and feedback the execution results based on the human-machine-object ternary knowledge graph, and update the knowledge graph according to the execution results.

8. An electronic device, characterized in that: include: Processor; as well as, A memory arranged to store computer-executable instructions, which, when executed, enable the processor to implement the steps of the knowledge graph-based smart community human-machine-object collaborative scheduling method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: Used to store computer executable instructions, which, when executed, implement the steps of the knowledge graph-based smart community human-machine-object collaborative scheduling method as described in any one of claims 1 to 6.

Images