Progressive Privacy Calculation Method and System for the Virtual-Reality Fusion Scenario of Intelligent Sweeping Robots

Abstract

The present invention provides a progressive privacy computing method and system for the virtual-real fusion scenario of an intelligent floor sweeper, which relates to the technical field of floor sweepers. It includes collecting indoor space data through multi-modal sensors, and constructing a scene knowledge graph through an entangled state registration network and a graph neural network. Then, the graph is hierarchically encrypted based on deep reinforcement learning and zero-knowledge proof, and different versions of scene data with different privacy protection intensities are generated through federated learning. Finally, path planning is carried out in a secure multi-party computing environment, and the cleaning strategy is optimized by using neuro-symbolic reasoning and graph attention networks. The present invention can realize the scene understanding and path planning of an intelligent floor sweeper, improve the cleaning efficiency and enhance the user experience on the premise of protecting user privacy.

Description

Technical Field

[0001] The present invention relates to the technology of floor cleaning machines, and particularly to a progressive privacy computing method and system for the virtual-real fusion scenario of intelligent floor cleaning machines. Background Art

[0002] The rapid development of smart home has spawned the demand for intelligent cleaning equipment. As a typical representative, floor cleaning robots are gradually becoming essential items in modern families. To improve cleaning efficiency and user experience, researchers are committed to developing intelligent floor cleaning robots that can understand and adapt to complex indoor environments. The existing floor cleaning robot technology mainly relies on SLAM (Simultaneous Localization and Mapping) technology, combined with sensors such as lidar and cameras for navigation and path planning. Some high-end products also introduce AI object recognition technology to achieve more intelligent cleaning strategies, such as avoiding obstacles and identifying room types.

[0003] Insufficient privacy protection: Existing floor cleaning robots usually collect a large amount of indoor environment data, including room layout, furniture placement, etc. These data may contain users' privacy information. The existing privacy protection mechanisms are usually relatively simple, such as data anonymization or encrypted storage, and it is difficult to effectively prevent the risks of data leakage and abuse.

[0004] Limited scene understanding ability: The current floor cleaning robots' understanding of the scene is mainly based on 2D maps and simple object recognition, lacking in-depth understanding of scene semantic information. For example, it is difficult for the robot to distinguish different room functional areas and identify the user's activity status, thus limiting its intelligence level and service ability.

[0005] Lack of flexibility in path planning: Most of the existing path planning algorithms are based on pre-set rules or simple optimization strategies, and it is difficult to adapt to complex and changeable indoor environments. For example, when the environment changes, the robot may need to reconstruct the map and plan the path again, with low efficiency. Moreover, the existing algorithms are difficult to customize cleaning strategies according to users' personalized needs. Summary of the Invention

[0006] The embodiments of the present invention provide a progressive privacy computing method and system for the virtual-real fusion scenario of intelligent floor cleaning machines, which can solve the problems in the prior art.

[0007] In the first aspect of the embodiments of the present invention,

[0008] A progressive privacy computing method for the virtual-real fusion scenario of intelligent floor cleaning machines is provided, including:

[0009] Collect indoor space data through the multi-modal sensor array of the intelligent sweeper, obtain the point cloud data of the micrometer-level lidar, the three-dimensional image data of the depth perception camera, the spatial response data of the sonar sensor, and the reflection signal data of the millimeter-wave radar, and input the point cloud data, the three-dimensional image data, the spatial response data, and the reflection signal data into the entangled state registration network to generate a six-dimensional scene feature tensor in a unified coordinate system; input the six-dimensional scene feature tensor into the graph neural network for semantic segmentation and temporal modeling to construct a scene knowledge graph including the spatial distribution of scene objects, object attribute information, and spatio-temporal interaction relationships between objects;

[0010] Input the scene knowledge graph into the deep reinforcement learning network, and construct a privacy sensitivity scoring matrix according to the object attribute information in the scene knowledge graph; based on the privacy sensitivity scoring matrix, use the zero-knowledge proof mechanism to perform hierarchical encryption processing on the nodes in the scene knowledge graph to generate a multi-level encrypted graph; submit the multi-level encrypted graph to the blockchain network node group, and calculate and output multiple scene data versions with different privacy protection intensities through the federated learning framework, and each scene data version includes corresponding privacy protection intensity parameters and scene information integrity parameters;

[0011] Construct a path planning optimization model in a secure multi-party computing environment according to the scene data version and its corresponding parameters; input the path planning optimization model into the neuro-symbolic reasoning engine to generate an initial set of cleaning path plans; use the graph attention network to perform distributed optimization evaluation on the initial set of cleaning path plans, and output the optimal cleaning strategy with a credibility quantification index; screen the optimal cleaning path based on the optimal cleaning strategy, and send the optimal cleaning path to the intelligent sweeper for execution through an encrypted channel; collect the real-time status data generated during the execution of the intelligent sweeper, and transmit the real-time status data back to the scene knowledge graph for dynamic update to achieve continuous optimization of the scene model.

[0012] Input the scene knowledge graph into the deep reinforcement learning network, and construct a privacy sensitivity scoring matrix according to the object attribute information in the scene knowledge graph; based on the privacy sensitivity scoring matrix, use the zero-knowledge proof mechanism to perform hierarchical encryption processing on the nodes in the scene knowledge graph to generate a multi-level encrypted graph, including:

[0013] Input the scene knowledge graph into a deep reinforcement learning network, which includes a state space module, an action space module, and a value evaluation module; the state space module receives the spatial position information, object usage frequency information, and object interaction relationship information of the objects in the scene knowledge graph; the action space module generates a privacy sensitivity score interval based on the output information of the state space module; the value evaluation module calculates a privacy sensitivity score value according to the spatio-temporal correlation of object attributes and user privacy annotation information;

[0014] Input the privacy sensitivity score value output by the deep reinforcement learning network into a scoring matrix construction module. The scoring matrix construction module uses an adaptive weight fusion algorithm to normalize the privacy sensitivity score value and generate a privacy sensitivity scoring matrix representing the privacy association degree between object nodes in the scene knowledge graph; based on the privacy sensitivity scoring matrix, use the spectral clustering algorithm to divide the object nodes into a sensitive protection level higher than a preset sensitive threshold and a sensitive protection level lower than the preset sensitive threshold;

[0015] Invoke a zero-knowledge proof mechanism for hierarchical encryption processing for object nodes of different sensitive protection levels. The zero-knowledge proof mechanism includes a proof generation unit and a verification execution unit; the proof generation unit constructs an encryption proof using the Groth16 protocol, adopts a fully homomorphic encryption scheme for object nodes of the sensitive protection level higher than the preset sensitive threshold, and adopts a lightweight searchable encryption scheme for object nodes of the sensitive protection level lower than the preset sensitive threshold; the verification execution unit verifies the legality of the encrypted data through pairing operations and outputs an encryption proof verification result, and reconstructs the encrypted object nodes that pass the verification into a multi-level encrypted graph.

[0016] Submit the multi-level encrypted graph to a blockchain network node group, and calculate and output multiple scene data versions with different privacy protection strengths through a federated learning framework. Each scene data version includes corresponding privacy protection strength parameters and scene information integrity parameters, including:

[0017] Slice the multi-level encrypted graph according to the spatial position relationship, and use the consistent hashing algorithm to distribute the data slices to multiple nodes in the blockchain network node group. The data slices include node encrypted attribute data, edge relationship encrypted data, and slice index information; in the blockchain network node group, use the practical Byzantine fault tolerance consensus algorithm to perform data consistency verification, calculation correctness verification, and version synchronization verification on the data slices;

[0018] Construct a federated learning framework based on the blockchain network node group. The federated learning framework includes a feature extraction network, a feature fusion network, and a decision network. The feature extraction network extracts features from the encrypted data on the local node. The feature fusion network integrates the feature information of multiple nodes. The decision network generates scenario data versions with different privacy protection intensities. In the federated learning framework, an adaptive federated optimization algorithm is used for model training. The adaptive federated optimization algorithm performs local model updates through nodes, reduces communication overhead via gradient compression, aggregates model parameters using homomorphic encryption, and introduces differential privacy to protect the gradient update process.

[0019] Input the output result of the federated learning framework into the scenario data generation module. The scenario data generation module generates privacy protection intensity parameters according to the data desensitization degree, encryption intensity, and noise level, and generates scenario information integrity parameters according to the topology preservation degree, attribute retention rate, and relationship integrity degree.

[0020] According to the scenario data version and its corresponding parameters, construct a path planning optimization model in a secure multi-party computing environment. Input the path planning optimization model into the neuro-symbolic reasoning engine to generate an initial set of cleaning path plans. Use a graph attention network to perform distributed optimization evaluation on the initial set of cleaning path plans and output an optimal cleaning strategy with a credibility quantification index, including:

[0021] According to the scenario data version and its corresponding parameters, construct a path planning optimization model in a secure multi-party computing environment. The path planning optimization model uses a three-party computing protocol based on secret sharing to decompose the computing tasks into spatial constraint computing tasks, temporal constraint computing tasks, and result aggregation computing tasks. The path planning optimization model outputs an initial solution for path planning that integrates physical constraint conditions, environmental constraint conditions, and privacy constraint conditions.

[0022] Input the initial solution for path planning into the neuro-symbolic reasoning engine for in-depth reasoning operations. The encoding layer of the neuro-symbolic reasoning engine converts the scenario space relationship, action temporal relationship, and constraint rule relationship in the initial solution for path planning into vector representations. The reasoning layer of the neuro-symbolic reasoning engine extracts implicit rules from historical data based on the vector representations and fuses them with expert rules to generate a path planning rule set. The decoding layer of the neuro-symbolic reasoning engine restores the reasoning result to a symbolic path description according to the path planning rule set to generate an initial set of cleaning path plans containing multiple candidate paths.

[0023] Construct a graph structure representation for the candidate paths in the initial cleaning path plan set, where the nodes of the graph structure representation correspond to path points and the edges correspond to path segments; input the graph structure representation into a graph attention network, and calculate node-level attention weights, edge-level attention weights, and global attention weights through distributed computing nodes; based on the node-level attention weights, the edge-level attention weights, and the global attention weights, perform priority sorting on the candidate paths to generate a sequence of path plans with evaluation scores;

[0024] Calculate a credibility quantization index for each path in the sequence of path plans, and input the credibility quantization index as an evaluation feedback into the graph attention network to update the node-level attention weights, the edge-level attention weights, and the global attention weights, optimize the cleaning of the sequence of path plans, and output an optimal cleaning strategy.

[0025] Inputting the graph structure representation into a graph attention network, calculating node-level attention weights, edge-level attention weights, and global attention weights through distributed computing nodes; performing priority sorting on candidate paths based on the node-level attention weights, the edge-level attention weights, and the global attention weights to generate a sequence of path plans with evaluation scores includes:

[0026] Obtain path features, where the path features include geometric features such as path segment length, path segment curvature, and path segment slope, as well as performance features such as energy consumption index and time index; construct the geometric features and the performance features into a node attribute vector, and construct the corner feature and the distance feature between path segments into an edge attribute vector;

[0027] Input the graph structure representation into a graph attention network, and process the node attribute vector and the edge attribute vector in parallel through distributed computing nodes; the distributed computing node constructs a query vector and a key vector based on the node attribute vector, and uses the self-attention mechanism to calculate the correlation between the query vector and the key vector to obtain the node attention weight;

[0028] The distributed computing node splices the node attribute vector and the edge attribute vector and inputs them into a trainable parameter matrix to calculate the constraint relationship between adjacent path segments to obtain the edge-level attention weight; the distributed computing node constructs a global context vector, splices the node attribute vector and the global context vector and inputs them into a multi-layer perceptron to calculate the relevance between the overall path feature and the local feature to obtain the global attention weight;

[0029] The distributed computing node performs weighted combination on the node attention weight, the edge-level attention weight, and the global attention weight based on a preset weight to obtain a path comprehensive score; sorts the candidate paths according to the path comprehensive score, and generates a path scheme sequence with evaluation scores.

[0030] Based on the optimal cleaning strategy, the optimal cleaning path is selected, and the optimal cleaning path is sent to the intelligent sweeper through an encrypted channel for execution; the real-time status data generated during the execution of the intelligent sweeper is collected, and the real-time status data is sent back to the scenario knowledge graph for dynamic update, and the continuous optimization of the scenario model includes:

[0031] Perform an execution feasibility assessment, a task completion assessment, and a resource consumption assessment on the path scheme based on the optimal cleaning strategy, and select the optimal cleaning path from the path scheme according to the assessment results; construct a secure transmission channel with end-to-end encryption function, and the secure transmission channel includes a hybrid encryption module, an identity authentication module, and a dynamic key update module; send the optimal cleaning path to the intelligent sweeper through the secure transmission channel, and the sending process uses a segmented confirmation and packet loss retransmission mechanism to ensure transmission reliability;

[0032] Collect real-time status data from the intelligent sweeper, preprocess the real-time status data, filter noise through Kalman filtering, realize multi-source data fusion through time series alignment, perform anomaly detection through statistical analysis, and generate preprocessed status data; send the preprocessed status data back to the scenario knowledge graph;

[0033] Dynamically update the scenario knowledge graph through time series knowledge fusion and spatial knowledge fusion. The time series knowledge fusion is processed based on historical data weight, new data credibility, and conflict data reconciliation. The spatial knowledge fusion is processed based on local update, global consistency, and topological optimization; optimize the motion model parameters, environment model parameters, and task model parameters based on the fusion results to achieve continuous optimization of the scenario model.

[0034] Perform an execution feasibility assessment, a task completion assessment, and a resource consumption assessment on the path scheme based on the optimal cleaning strategy, and select the optimal cleaning path from the path scheme according to the assessment results; construct a secure transmission channel with end-to-end encryption function, including:

[0035] Perform grid division on the cleaning area based on the optimal cleaning strategy, and generate a path plan to be evaluated; perform an execution feasibility assessment based on the optimal cleaning strategy, perform Bezier curve fitting on the path plan to be evaluated to obtain path curvature data, establish a kinematic model based on the path curvature data to calculate the maximum allowable speed of each path segment, analyze the attitude stability margin using the zero moment point theory according to the maximum allowable speed, establish a wheel-ground force model in combination with the attitude stability margin and ground characteristics, verify the distance constraint between adjacent path points and speed continuity, and generate an execution feasibility assessment result;

[0036] Perform a task completion assessment based on the optimal cleaning strategy, calculate the coverage area of the sweeper in the grid-divided cleaning area according to the path plan to be evaluated, count the number of effectively covered grids, detect the multiple coverage areas and their time distributions, identify the uncovered areas using connected component analysis and evaluate their accessibility, generate a supplementary cleaning path for the uncovered areas, and obtain a task completion assessment result based on the supplementary cleaning path;

[0037] Perform a resource consumption assessment based on the optimal cleaning strategy, calculate the dynamic power consumption according to the path curvature data and the maximum allowable speed, calculate the operation time based on the coverage area, evaluate the obstacle avoidance time in combination with the wheel-ground force model, analyze the complexity of the path planning algorithm and the real-time control calculation amount, and generate a resource consumption assessment result;

[0038] According to the execution feasibility assessment result, the task completion assessment result, and the resource consumption assessment result, screen the optimal cleaning path from the path plan to be evaluated; construct a secure transmission channel with end-to-end encryption function based on the selected optimal cleaning path.

[0039] In the second aspect of the embodiments of the present invention,

[0040] Provide a progressive privacy computing system for the virtual-real fusion scenario of an intelligent sweeper, including:

[0041] The first unit is used to collect indoor space data through the multi-modal sensor array of the intelligent sweeper, obtain the point cloud data of the micron-level lidar, the three-dimensional image data of the depth perception camera, the spatial response data of the sonar sensor, and the reflection signal data of the millimeter-wave radar, and input the point cloud data, the three-dimensional image data, the spatial response data, and the reflection signal data into the entangled state registration network to generate a six-dimensional scene feature tensor in a unified coordinate system; input the six-dimensional scene feature tensor into the graph neural network for semantic segmentation and temporal modeling, and construct a scene knowledge graph including the spatial distribution of scene objects, object attribute information, and spatio-temporal interaction relationships between objects;

[0042] A second unit, configured to input the scene knowledge graph into a deep reinforcement learning network, construct a privacy sensitivity scoring matrix according to the object attribute information in the scene knowledge graph; based on the privacy sensitivity scoring matrix, use a zero-knowledge proof mechanism to perform hierarchical encryption processing on the nodes in the scene knowledge graph to generate a multi-level encrypted graph; submit the multi-level encrypted graph to a blockchain network node group, and calculate and output multiple scene data versions with different privacy protection strengths through a federated learning framework, and each scene data version includes corresponding privacy protection strength parameters and scene information integrity parameters;

[0043] A third unit, configured to construct a path planning optimization model in a secure multi-party computing environment according to the scene data version and its corresponding parameters; input the path planning optimization model into a neuro-symbolic reasoning engine to generate an initial set of sweeping path plans; use a graph attention network to perform distributed optimization evaluation on the initial set of sweeping path plans, and output an optimal sweeping strategy with a credibility quantification index; screen the optimal sweeping path based on the optimal sweeping strategy, and send the optimal sweeping path to the intelligent sweeper for execution through an encrypted channel; collect the real-time status data generated by the intelligent sweeper during the execution process, and send the real-time status data back to the scene knowledge graph for dynamic update to achieve continuous optimization of the scene model.

[0044] In the third aspect of the embodiments of the present invention,

[0045] There is provided an electronic device, including:

[0046] A processor;

[0047] A memory for storing instructions executable by the processor;

[0048] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0049] In the fourth aspect of the embodiments of the present invention,

[0050] There is provided a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0051] The beneficial effects of this application are as follows:

[0052] 1. Improve scene understanding ability: By fusing multi-modal sensor data and graph neural networks, a scene knowledge graph including the spatial distribution of objects, attribute information, and spatio-temporal interaction relationships is constructed, enabling the sweeper to more comprehensively understand the indoor environment and improving the cleaning efficiency and intelligence level.

[0053] 2. Ensure user privacy and security: Process scenario data using zero-knowledge proofs and multi-level encryption technologies, and combine federated learning and secure multi-party computation to achieve the output of scenario data versions under different privacy protection intensities, effectively protecting user privacy while sharing and utilizing data.

[0054] 3. Optimize the cleaning path planning: Based on technologies such as deep reinforcement learning, neuro-symbolic reasoning, and graph attention networks, construct a path planning optimization model, and dynamically update the scenario model in combination with real-time status data to achieve a more intelligent, efficient, and reliable cleaning path planning. Brief Description of the Drawings

[0055] Figure 1 It is a schematic flowchart of the progressive privacy calculation method for the virtual-real fusion scenario of the intelligent sweeper in the embodiment of the present invention;

[0056] Figure 2 It is a schematic structural diagram of the progressive privacy calculation system for the virtual-real fusion scenario of the intelligent sweeper in the embodiment of the present invention. Detailed Embodiments

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0058] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0059] Figure 1 It is a schematic flowchart of the progressive privacy calculation method for the virtual-real fusion scenario of the intelligent sweeper in the embodiment of the present invention, as Figure 1 shown, the method includes:

[0060] S11. Collect indoor space data through the multi-modal sensor array of the intelligent sweeper, obtain the point cloud data of the micrometer-level lidar, the three-dimensional image data of the depth perception camera, the spatial response data of the sonar sensor, and the reflection signal data of the millimeter-wave radar, and input the point cloud data, the three-dimensional image data, the spatial response data, and the reflection signal data into the entangled state registration network to generate a six-dimensional scene feature tensor in a unified coordinate system; input the six-dimensional scene feature tensor into the graph neural network for semantic segmentation and temporal modeling to construct a scene knowledge graph including the spatial distribution of scene objects, object attribute information, and spatio-temporal interaction relationships between objects;

[0061] S12. Input the scenario knowledge graph into a deep reinforcement learning network, construct a privacy sensitivity scoring matrix according to the object attribute information in the scenario knowledge graph; based on the privacy sensitivity scoring matrix, use the zero-knowledge proof mechanism to perform hierarchical encryption processing on the nodes in the scenario knowledge graph to generate a multi-level encrypted graph; submit the multi-level encrypted graph to the blockchain network node group, calculate and output multiple scenario data versions with different privacy protection strengths through the federated learning framework, and each scenario data version includes corresponding privacy protection strength parameters and scenario information integrity parameters;

[0062] S13. According to the scenario data version and its corresponding parameters, construct a path planning optimization model in a secure multi-party computing environment; input the path planning optimization model into a neuro-symbolic inference engine to generate an initial set of cleaning path plans; use a graph attention network to perform distributed optimization evaluation on the initial set of cleaning path plans, and output an optimal cleaning strategy with a credibility quantification index; based on the optimal cleaning strategy, select the optimal cleaning path, and send the optimal cleaning path to the intelligent sweeper for execution through an encrypted channel; collect the real-time status data generated during the execution of the intelligent sweeper, and send the real-time status data back to the scenario knowledge graph for dynamic update to achieve continuous optimization of the scenario model.

[0063] In an alternative embodiment, input the scenario knowledge graph into a deep reinforcement learning network, construct a privacy sensitivity scoring matrix according to the object attribute information in the scenario knowledge graph; based on the privacy sensitivity scoring matrix, use the zero-knowledge proof mechanism to perform hierarchical encryption processing on the nodes in the scenario knowledge graph to generate a multi-level encrypted graph, including:

[0064] Input the scenario knowledge graph into a deep reinforcement learning network, where the deep reinforcement learning network includes a state space module, an action space module, and a value evaluation module; the state space module receives the spatial position information, object usage frequency information, and object interaction relationship information of the objects in the scenario knowledge graph; the action space module generates a privacy sensitivity scoring interval based on the output information of the state space module; the value evaluation module calculates the privacy sensitivity scoring value according to the spatio-temporal correlation of object attributes and user privacy annotation information;

[0065] Input the privacy sensitivity score value output by the deep reinforcement learning network into a scoring matrix construction module. The scoring matrix construction module uses an adaptive weight fusion algorithm to normalize the privacy sensitivity score value, generating a privacy sensitivity scoring matrix that characterizes the privacy association degree between object nodes in the scenario knowledge graph. Based on the privacy sensitivity scoring matrix, use the spectral clustering algorithm to divide the object nodes into a sensitive protection level higher than a preset sensitive threshold and a sensitive protection level lower than the preset sensitive threshold.

[0066] Invoke a zero-knowledge proof mechanism for hierarchical encryption processing for object nodes of different sensitive protection levels. The zero-knowledge proof mechanism includes a proof generation unit and a verification execution unit. The proof generation unit constructs an encryption proof using the Groth16 protocol, adopts a fully homomorphic encryption scheme for object nodes of the sensitive protection level higher than the preset sensitive threshold, and adopts a lightweight searchable encryption scheme for object nodes of the sensitive protection level lower than the preset sensitive threshold. The verification execution unit verifies the legality of the encrypted data through pairing operations, outputs the encryption proof verification result, and reconstructs the encrypted object nodes that pass the verification into a multi-level encrypted graph.

[0067] First, construct a scenario knowledge graph. The scenario knowledge graph is a graph structure composed of nodes and edges. The nodes represent objects in the scenario, and the edges represent the relationships between objects. Each node contains attribute information of the object, such as: name, location, usage frequency, etc. For example, in the knowledge graph of a smart home scenario, the nodes can be "smart speaker", "camera", "TV", etc., and the edges can be "control", "connect", "located in the same room", etc.

[0068] Next, input the scenario knowledge graph into a deep reinforcement learning network. This network includes a state space module, an action space module, and a value evaluation module. The state space module receives the spatial position information, usage frequency information of objects in the scenario knowledge graph, and the interaction relationship information between objects. For example, the location, usage frequency of the smart speaker, and its connection relationship with other devices. The action space module generates a privacy sensitivity score interval based on the output information of the state space module. For example, between 0 and 1, divided into three intervals: low, medium, and high. The value evaluation module calculates the privacy sensitivity score value according to the spatio-temporal correlation of object attributes and user privacy annotation information. For example, if the camera is labeled as "highly sensitive" by the user, its privacy sensitivity score value will be higher. Suppose the usage frequency of the smart speaker is high and it has a connection relationship with the camera, then its privacy sensitivity score value will also increase accordingly. Suppose the finally calculated score value is 0.8.

[0069] Then, construct a privacy sensitivity scoring matrix. Input the privacy sensitivity scoring values output by the deep reinforcement learning network into the scoring matrix construction module. This module uses an adaptive weight fusion algorithm to normalize the privacy sensitivity scoring values and generate a privacy sensitivity scoring matrix that represents the privacy association degree between object nodes in the scenario knowledge graph. For example, if the privacy sensitivity scoring values of a smart speaker and a camera are 0.8 and 0.9 respectively, the corresponding element values in the matrix will be relatively high, indicating a strong privacy association between them.

[0070] Subsequently, classify the object nodes. Based on the privacy sensitivity scoring matrix, use the spectral clustering algorithm to divide the object nodes into a sensitive protection level higher than the preset sensitive threshold and a sensitive protection level lower than the preset sensitive threshold. Assume the preset sensitive threshold is 0.7, then both the smart speaker and the camera belong to the sensitive protection level higher than the preset sensitive threshold.

[0071] Finally, perform hierarchical encryption processing on the nodes. Invoke the zero-knowledge proof mechanism for hierarchical encryption processing for object nodes of different sensitive protection levels. The zero-knowledge proof mechanism includes a proof generation unit and a verification execution unit. The proof generation unit constructs an encryption proof using the Groth16 protocol, adopts a fully homomorphic encryption scheme for object nodes of the sensitive protection level higher than the preset sensitive threshold, and adopts a lightweight searchable encryption scheme for object nodes of the sensitive protection level lower than the preset sensitive threshold. For example, the smart speaker and the camera are encrypted using the fully homomorphic encryption scheme, while other nodes with lower sensitivity are encrypted using the lightweight searchable encryption scheme. The verification execution unit verifies the legality of the encrypted data through pairing operations and outputs the verification result of the encryption proof, and reconstructs the encrypted object nodes into a multi-level encrypted graph.

[0072] The solution of this application can:

[0073] Enhance the privacy protection ability: Through deep reinforcement learning and zero-knowledge proof technologies, fine-grained hierarchical encryption of sensitive information in the scenario knowledge graph is achieved, effectively preventing privacy leakage and ensuring data security. Improve data availability: Adopt a hierarchical encryption strategy. While protecting sensitive information, it allows access to and utilization of low-sensitivity data, ensuring data availability and avoiding data access restrictions brought by the "one-size-fits-all" encryption method. Improve system efficiency: The application of the lightweight searchable encryption scheme reduces the computational overhead of encryption and decryption, improves the overall efficiency of the system, and makes this solution more practical.

[0074] In an alternative embodiment, the multi-level encrypted graph is submitted to a blockchain network node group, and multiple scenario data versions with different privacy protection strengths are calculated and output through a federated learning framework. Each scenario data version includes corresponding privacy protection strength parameters and scenario information integrity parameters, including:

[0075] The multi-level encrypted graph is fragmented according to the spatial position relationship, and the data fragments are allocated to multiple nodes in the blockchain network node group by using the consistent hashing algorithm. The data fragments include node encryption attribute data, edge relationship encryption data, and fragment index information; in the blockchain network node group, the practical Byzantine fault tolerance consensus algorithm is used to perform data consistency verification, calculation correctness verification, and version synchronization verification on the data fragments;

[0076] A federated learning framework is constructed based on the blockchain network node group. The federated learning framework includes a feature extraction network, a feature fusion network, and a decision network; the feature extraction network extracts features from the encrypted data of the nodes locally, the feature fusion network integrates the feature information of multiple nodes, and the decision network generates scenario data versions with different privacy protection strengths; in the federated learning framework, the adaptive federated optimization algorithm is used for model training. The adaptive federated optimization algorithm performs local model updates through nodes, reduces communication overhead through gradient compression, aggregates model parameters using homomorphic encryption, and introduces differential privacy protection for the gradient update process;

[0077] The output result of the federated learning framework is input into a scenario data generation module, and the scenario data generation module generates privacy protection strength parameters according to the data desensitization degree, encryption strength, and noise level, and generates scenario information integrity parameters according to the topology preservation degree, attribute retention rate, and relationship integrity degree.

[0078] First, prepare the multi-level encrypted graph data. The graph includes nodes and edges. The nodes have attribute information, and the edges represent the relationships between the nodes. All sensitive data is encrypted. For example, node attributes can use attribute encryption, and edge relationships can use homomorphic encryption. The graph data is fragmented according to the spatial position. Each fragment includes node encryption attribute data, edge relationship encryption data, and fragment index information. For example, a city traffic graph can be divided into multiple fragments according to administrative regions, and each fragment includes information such as roads and intersections within the administrative region, as well as corresponding encrypted attribute and relationship data.

[0079] Then, distribute the data shards to the blockchain network node group. The consistent hashing algorithm is used to allocate each data shard to a suitable node, ensuring balanced data distribution and easy lookup. For example, the SHA256 algorithm is used to calculate the hash value of each shard, and then the shards are mapped to different blockchain nodes according to the hash values. The blockchain network adopts the Practical Byzantine Fault Tolerance (PBFT) consensus algorithm to perform data consistency verification, calculation correctness verification, and version synchronization verification on the data shards, ensuring that the data of all nodes is consistent, secure, and reliable. For example, when a node receives a new data shard, other nodes will jointly verify the integrity and correctness of the shard to prevent malicious nodes from tampering with the data.

[0080] Next, construct a federated learning framework based on the blockchain network node group. This framework includes a feature extraction network, a feature fusion network, and a decision network. The feature extraction network extracts features from the encrypted data locally at each node. For example, a convolutional neural network is used to extract road traffic flow features. The feature fusion network integrates the feature information extracted by multiple nodes. For example, a secure multi-party computation protocol is used to perform weighted averaging on the feature vectors of each node without revealing the original data. The decision network generates scenario data versions with different privacy protection strengths based on the fused feature information. For example, by adjusting the output layer parameters of the decision network, the desensitization degree of the generated data can be controlled. In the federated learning framework, the Adaptive Federated Optimization (AFO) algorithm is used for model training. This algorithm performs local model updates at the nodes and uses gradient compression to reduce communication overhead. At the same time, homomorphic encryption is used for model parameter aggregation, and differential privacy mechanisms are introduced to protect the gradient update process, further enhancing the privacy protection ability. For example, after each node locally trains the model, only the encrypted gradient information is uploaded to the central server, and the server aggregates the encrypted gradients and then distributes them to each node.

[0081] Finally, input the output result of the federated learning framework into the scenario data generation module. This module generates privacy protection strength parameters based on the data desensitization degree, encryption strength, and noise level, and generates scenario information integrity parameters based on the topology preservation degree, attribute retention rate, and relationship integrity degree. For example, if the data desensitization degree is high, the privacy protection strength parameter is also correspondingly high; if the topological structure is kept intact, the topology preservation degree parameter is high. Finally, multiple scenario data versions with different privacy protection strengths are output, and each version contains the corresponding privacy protection strength parameter and scenario information integrity parameter. For example, a highly desensitized data version can be generated for public release; at the same time, a lowly desensitized data version can be generated for internal research.

[0082] The solution of this application can:

[0083] Enhanced Data Privacy Protection: Through the combination of technologies such as multi-level encryption, blockchain technology, federated learning, and differential privacy, comprehensive protection of sensitive data is achieved, effectively preventing data leakage and abuse. Improve Data Availability: While protecting data privacy, this solution can generate multiple scenario data versions with different privacy protection intensities to meet data usage requirements in different scenarios, improving data availability and value. Ensure Data Integrity and Consistency: Utilizing the distributed ledger and consensus mechanism of blockchain, data integrity and consistency are ensured, preventing data from being tampered with and forged, and enhancing data credibility.

[0084] In an alternative implementation, according to the scenario data version and its corresponding parameters, a path planning optimization model is constructed in a secure multi-party computing environment; the path planning optimization model is input into a neuro-symbolic reasoning engine to generate an initial set of cleaning path plans; the initial set of cleaning path plans is evaluated and optimized distributively using a graph attention network, and the output optimal cleaning strategy with a credibility quantification index includes:

[0085] According to the scenario data version and its corresponding parameters, a path planning optimization model is constructed in a secure multi-party computing environment; the path planning optimization model adopts a three-party computing protocol based on secret sharing, decomposing the computing tasks into spatial constraint computing tasks, temporal constraint computing tasks, and result aggregation computing tasks; the path planning optimization model outputs an initial solution for path planning that integrates physical constraint conditions, environmental constraint conditions, and privacy constraint conditions.

[0086] The initial solution for path planning is input into a neuro-symbolic reasoning engine for in-depth reasoning operations. The encoding layer of the neuro-symbolic reasoning engine converts the scenario space relationship, action temporal relationship, and constraint rule relationship in the initial solution for path planning into vector representations; the reasoning layer of the neuro-symbolic reasoning engine extracts implicit rules from historical data based on the vector representations and fuses them with expert rules to generate a path planning rule set; the decoding layer of the neuro-symbolic reasoning engine restores the reasoning results to symbolic path descriptions according to the path planning rule set, generating an initial set of cleaning path plans containing multiple candidate paths.

[0087] A graph structure representation is constructed for the candidate paths in the initial set of cleaning path plans. The nodes of the graph structure representation correspond to path points, and the edges correspond to path segments; the graph structure representation is input into a graph attention network, and node-level attention weights, edge-level attention weights, and global attention weights are calculated through distributed computing nodes; based on the node-level attention weights, the edge-level attention weights, and the global attention weights, the candidate paths are sorted by priority, generating a sequence of path plans with evaluation scores.

[0088] Calculate the credibility quantization index for each path in the path plan sequence, and use the credibility quantization index as the evaluation feedback to input into the graph attention network, update the node-level attention weight, the edge-level attention weight, and the global attention weight, optimize the cleaning of the path plan sequence, and output the optimal cleaning strategy.

[0089] First, prepare the scenario data. The scenario data includes static data and dynamic data. The static data includes, for example, the room layout, furniture positions, etc. Taking version number v1.0 as an example, the parameters include the room size (10 meters long, 8 meters wide), the obstacle positions (coordinates (2,3), size 1x1 meter, coordinates (5,6), size 2x2 meters), etc. The dynamic data includes, for example, the real-time positions of people, pet positions, etc. Taking version number v2.0 as an example, the parameters include the person position (coordinates (3,4)), the pet position (coordinates (7,2)), etc. These data will be used to construct the path planning model.

[0090] Then, construct a path planning optimization model in the secure multi-party computing environment. Adopt a three-party computing protocol based on secret sharing, and decompose the computing tasks into spatial constraint computing tasks, temporal constraint computing tasks, and result aggregation computing tasks. Taking three computing nodes A, B, and C as an example, the room size information is secretly shared among the three nodes. Node A holds the first half value 5 of the room length, node B holds the second half value 5 of the room length, and node C holds the key for decryption. The three nodes respectively calculate the relationship between the data they hold and the obstacle positions, and secretly share and aggregate the intermediate results. Finally, an initial solution for path planning that integrates physical constraint conditions (such as room boundaries, obstacle positions), environmental constraint conditions (such as avoiding the positions of people), and privacy constraint conditions (such as not leaking the position information of people) is obtained. Assume the initial solution is a series of discrete coordinate points [(1,1), (2,2), (3,3), (4,4)].

[0091] Next, input the initial solution of path planning into the neuro-symbolic reasoning engine. The encoding layer of the neuro-symbolic reasoning engine converts the scene spatial relationships (e.g., the coordinate points (1,1) and (2,2) are adjacent), action temporal relationships (e.g., the order from (1,1) to (2,2) and then to (3,3)), and constraint rule relationships (e.g., avoiding obstacles) in the initial solution of path planning into vector representations. For example, represent the coordinate point (1,1) as the vector [0.1, 0.1], (2,2) as the vector [0.2, 0.2], the adjacent relationship as the vector [0.9], and the rule of avoiding obstacles as the vector [0.8]. Based on these vector representations, the reasoning layer of the neuro-symbolic reasoning engine extracts implicit rules (e.g., usually clean the corners of the room first) from the historical cleaning data and fuses them with expert rules (e.g., clean the high-frequency activity areas first) to generate a path planning rule set. For example, the rule set can be expressed as "clean the corners first, then the high-frequency areas, and finally the other areas". The decoding layer of the neuro-symbolic reasoning engine restores the reasoning result to a symbolic path description according to the path planning rule set, such as "from (1,1) to (4,1), and then from (4,1) to (4,4)", and generates an initial cleaning path plan set containing multiple candidate paths, such as {Path 1: [(1,1), (2,1), (3,1), (4,1), (4,2), (4,3), (4,4)], Path 2: [(1,1), (1,2), (1,3), (1,4), (2,4),(3,4), (4,4)]}.

[0092] Subsequently, construct a graph structure representation for the candidate paths in the initial cleaning path plan set. The nodes of the graph structure representation correspond to path points, and the edges correspond to path segments. Input the graph structure representation into the graph attention network, and calculate the node-level attention weight, edge-level attention weight, and global attention weight through distributed computing nodes. For example, the node-level attention weight of the path point (1,1) is 0.2, the edge-level attention weight of the edge (1,1)-(2,1) is 0.3, and the global attention weight of Path 1 is 0.5. Sort the candidate paths according to these weights to generate a sequence of path plans with evaluation scores. For example, if the evaluation score of Path 1 is 0.8 and the evaluation score of Path 2 is 0.7, then the sequence of path plans is {Path 1, Path 2}.

[0093] Finally, calculate the credibility quantization index for each path in the sequence of path plans. For example, the credibility of Path 1 is 0.9 and the credibility of Path 2 is 0.8. Input the credibility quantization index as an evaluation feedback into the graph attention network to update the node-level attention weight, edge-level attention weight, and global attention weight, optimize the sequence of path plans, and output the optimal cleaning strategy, such as selecting Path 1 as the optimal cleaning strategy.

[0094] The solution of this application can:

[0095] Privacy protection: Path planning is carried out in a secure multi-party computing environment, protecting privacy information in the scenario data, such as personnel locations, etc. Intelligent optimization: Using a neuro-symbolic reasoning engine and a graph attention network, it can intelligently generate and optimize the cleaning path according to the scenario data and constraints, improving the cleaning efficiency. Credibility quantification: By calculating the credibility quantification index, the reliability of the path plan can be evaluated, and the optimal cleaning strategy can be selected to improve the success rate of cleaning.

[0096] In an alternative embodiment, the graph structure representation is input into the graph attention network, and the node-level attention weight, edge-level attention weight, and global attention weight are calculated by a distributed computing node; based on the node-level attention weight, the edge-level attention weight, and the global attention weight, the candidate paths are prioritized, and the generation of a sequence of path plans with evaluation scores includes:

[0097] Obtain path features, where the path features include geometric features such as path segment length, path segment curvature, and path segment slope, as well as performance features such as energy consumption index and time index; construct the geometric features and the performance features into a node attribute vector, and construct the corner feature and distance feature between path segments into an edge attribute vector;

[0098] Input the graph structure representation into the graph attention network, and the distributed computing node processes the node attribute vector and the edge attribute vector in parallel; the distributed computing node constructs a query vector and a key vector based on the node attribute vector, and uses the self-attention mechanism to calculate the correlation between the query vector and the key vector to obtain the node attention weight;

[0099] The distributed computing node splices the node attribute vector and the edge attribute vector and inputs them into a trainable parameter matrix to calculate the constraint relationship between adjacent path segments to obtain the edge-level attention weight; the distributed computing node constructs a global context vector, splices the node attribute vector and the global context vector and inputs them into a multi-layer perceptron to calculate the correlation between the overall path feature and the local feature to obtain the global attention weight;

[0100] The distributed computing node performs a weighted combination of the node attention weight, the edge-level attention weight, and the global attention weight based on a preset weight to obtain a comprehensive path score; according to the comprehensive path score, the candidate paths are prioritized to generate a sequence of path plans with evaluation scores.

[0101] First, obtain the features of each candidate path. These features include the path segment length, path segment curvature, and path segment slope that describe the path geometry, as well as the energy consumption index and time index that reflect the path performance.

[0102] Next, the extracted path features are constructed into the input of a graph structure representation. Each path segment is regarded as a node in the graph, and the geometric features and performance features of the path segment constitute the attribute vector of the node. The turning angle features and distance features between path segments constitute the attribute vector of the edge connecting these nodes. For example, if the turning angle between path segment 1 and path segment 2 is 30 degrees and the distance is 50 meters, then the edge attribute vector between them is (30, 50).

[0103] Then, the constructed graph structure representation is input into a graph attention network for processing. The distributed computing nodes in the network process the node attribute vectors and edge attribute vectors in parallel. Each distributed computing node first constructs a query vector and a key vector based on the node attribute vector, and obtains the node attention weight by calculating the correlation between the query vector and the key vector, which reflects the importance of each path segment itself.

[0104] At the same time, each distributed computing node concatenates the node attribute vector and the edge attribute vector, and then inputs them into a trainable parameter matrix to calculate the constraint relationship between adjacent path segments, obtaining the edge-level attention weight, which reflects the mutual influence between path segments.

[0105] In addition, each distributed computing node also constructs a global context vector, concatenates the node attribute vector and the global context vector and inputs them into a multi-layer perceptron to calculate the correlation between the overall path feature and the local feature, obtaining the global attention weight, which reflects the influence of each path segment on the overall path.

[0106] After that, each distributed computing node performs a weighted combination of the node attention weight, the edge-level attention weight, and the global attention weight according to the preset weights to obtain the comprehensive score of each path. For example, if the preset weights are 0.4, 0.3, 0.3, then the comprehensive score of a path is: 0.4 * node attention weight + 0.3 * edge-level attention weight + 0.3 * global attention weight.

[0107] Finally, the candidate paths are sorted by priority according to the calculated path comprehensive scores, generating a sequence of path schemes with evaluation scores. For example, if the comprehensive scores of three candidate paths are 85, 92, and 78 respectively, then the generated sequence of path schemes is: Path 2 (92 points), Path 1 (85 points), Path 3 (78 points).

[0108] The solution of this application can:[[]]

[0109] By comprehensively considering the geometric features, performance features of the path, and the relationships between path segments, the advantages and disadvantages of the path can be evaluated more comprehensively, thus enabling the selection of a more suitable path. Using distributed computing nodes to process information in parallel improves the efficiency of path planning, especially suitable for dealing with large-scale road networks and complex scenarios. Utilizing the learning ability of the graph attention network, the model parameters can be adjusted according to different application scenarios and requirements, thereby achieving more accurate path planning.

[0110] In an alternative embodiment, the optimal cleaning path is screened based on the optimal cleaning strategy, and the optimal cleaning path is sent to the intelligent sweeper for execution through an encrypted channel; the real-time status data generated during the execution of the intelligent sweeper is collected, and the real-time status data is sent back to the scenario knowledge graph for dynamic update, and the continuous optimization of the scenario model includes:

[0111] Perform an execution feasibility assessment, task completion assessment, and resource consumption assessment on the path plan based on the optimal cleaning strategy, and screen the optimal cleaning path from the path plan according to the assessment results; construct a secure transmission channel with end-to-end encryption function, and the secure transmission channel includes a hybrid encryption module, an identity authentication module, and a dynamic key update module; send the optimal cleaning path to the intelligent sweeper through the secure transmission channel, and the transmission reliability is ensured by using a segmented confirmation and packet loss retransmission mechanism during the sending process;

[0112] Collect real-time status data from the intelligent sweeper, preprocess the real-time status data, filter the noise through Kalman filtering, achieve multi-source data fusion through time series alignment, perform anomaly detection through statistical analysis, and generate preprocessed status data; send the preprocessed status data back to the scenario knowledge graph;

[0113] Dynamically update the scenario knowledge graph through time series knowledge fusion and spatial knowledge fusion. The time series knowledge fusion is processed based on historical data weights, new data credibility, and conflict data reconciliation. The spatial knowledge fusion is processed based on local update, global consistency, and topological optimization; optimize the motion model parameters, environment model parameters, and task model parameters based on the fusion results to achieve the continuous optimization of the scenario model.

[0114] First, construct a scenario knowledge graph. This graph contains static data such as the geometric structure of the room, furniture positions, obstacle information, etc., and dynamic data such as dust distribution, floor materials, etc. For example, a room can be represented as a two-dimensional floor plan, furniture and obstacles are represented by polygons, dust distribution is represented by grayscale values, and floor materials are represented by different labels. The initial graph can be constructed by manual input from the user or sensor scanning.

[0115] Next, several candidate cleaning path plans are generated based on the optimal cleaning strategy. The optimal cleaning strategy can be set according to user requirements. For example, areas with more dust can be preferentially cleaned, or high-frequency activity areas can be preferentially cleaned. Taking the "zigzag" cleaning as an example, multiple different "zigzag" paths can be generated according to the geometric shape of the room and the furniture layout. Suppose the room is a 5-meter by 4-meter rectangle, two candidate paths can be generated: one from left to right and one from top to bottom.

[0116] Then, the candidate path plans are evaluated. The evaluation metrics include execution feasibility, task completion rate, and resource consumption. Execution feasibility refers to whether the path will collide with obstacles; task completion rate refers to the proportion of the cleaning area covered by the path; resource consumption refers to the estimated time and power required to complete the path. For example, suppose the path from left to right will collide with a table, then its execution feasibility is low; the path from top to bottom can cover 90% of the area, with an estimated time of 20 minutes and a power consumption of 20%, then its task completion rate is high and the resource consumption is moderate.

[0117] The optimal cleaning path is selected according to the evaluation results. The path with the highest comprehensive score of the evaluation metrics is selected as the optimal path. For example, if the path from top to bottom has the highest comprehensive score in all metrics, then this path is selected as the optimal cleaning path.

[0118] A secure communication channel is established to send the optimal cleaning path to the intelligent floor sweeper. This channel adopts a hybrid encryption mechanism, combining symmetric encryption and asymmetric encryption, to ensure the security of data transmission. First, the floor sweeper and the control center exchange symmetric keys through an asymmetric encryption algorithm; then, the symmetric key is used to encrypt the path data for transmission. At the same time, the channel also includes an identity authentication module and a dynamic key update module to further enhance security. For example, the floor sweeper and the control center can pre-share a public key and private key pair for identity authentication and key exchange; every once in a while, the symmetric key is dynamically updated to prevent key leakage. During the transmission process, a segmented confirmation and packet loss retransmission mechanism is adopted to ensure that the path data is transmitted to the floor sweeper completely and reliably. For example, the path data is divided into multiple data packets, and after each data packet is sent, the floor sweeper needs to return a confirmation message; if the control center does not receive the confirmation message, the data packet is re-sent.

[0119] The intelligent floor sweeper starts to execute the cleaning task and real-time collects status data, including position coordinates, power, sensor data, etc. For example, the floor sweeper uploads the current position coordinates, remaining power, and dust sensor readings once per second.

[0120] Preprocess the collected status data. First, use the Kalman filter algorithm to filter the noise in the data. For example, remove the jitter noise in the position coordinates. Then, fuse the data from different sensors through time series alignment technology. For example, associate the position coordinates and the readings of the dust sensor to the same time point. Finally, perform anomaly detection through statistical analysis methods. For example, determine whether a sudden drop in power is an abnormal situation.

[0121] Transmit the preprocessed status data back to the scenario knowledge graph for dynamic update. For example, mark the areas cleaned by the sweeper as cleaned, update the dust distribution information, and record the position information of the obstacles.

[0122] The update of the scenario knowledge graph includes temporal knowledge fusion and spatial knowledge fusion. Temporal knowledge fusion considers the weights of historical data, the credibility of new data, and the reconciliation of conflicting data. For example, if there is a conflict between the new dust distribution information and the historical data, make a trade-off based on the reliability of the data sources. Spatial knowledge fusion considers local updates, global consistency, and topological optimization. For example, after updating the dust distribution information in a certain area, check whether it is consistent with the information in the surrounding areas and perform topological optimization on the map to maintain the connectivity and integrity of the map.

[0123] Finally, based on the fused scenario knowledge graph, optimize the parameters of the scenario model. For example, adjust the weight parameters of the cleaning strategy according to the updated dust distribution information; optimize the parameters of the path planning algorithm according to the position information of the obstacles.

[0124] The solution of this application can:

[0125] Improve the cleaning efficiency: Through dynamic path planning and self-learning of the scenario model, the sweeper can more effectively cover the cleaning area, reduce repeated cleaning and missed cleaning, thereby improving the cleaning efficiency. Enhance the environmental adaptability: By continuously updating the scenario knowledge graph, the sweeper can better adapt to environmental changes. For example, changes in the positions of furniture, the emergence of new obstacles, etc., thereby maintaining the stability of the cleaning effect. Improve the intelligent level: Through automatic path planning and self-learning of the scenario model, without manual intervention, the sweeper can autonomously complete the cleaning task, improving the intelligent level and enhancing the user experience.

[0126] In an alternative embodiment, perform feasibility evaluation, task completion evaluation, and resource consumption evaluation on the path plan based on the optimal cleaning strategy, and screen the optimal cleaning path from the path plan; constructing a secure transmission channel with end-to-end encryption function includes:

[0127] The cleaning area is rasterized based on the optimal cleaning strategy, and a path plan to be evaluated is generated; the execution feasibility is evaluated based on the optimal cleaning strategy. Bezier curve fitting is performed on the path plan to be evaluated to obtain path curvature data. A kinematic model is established based on the path curvature data to calculate the maximum allowable speed of each path segment. The zero moment point theory is used to analyze the attitude stability margin according to the maximum allowable speed. A wheel-ground force model is established by combining the attitude stability margin and the ground characteristics to verify the distance constraint between adjacent path points and the speed continuity, and an execution feasibility evaluation result is generated;

[0128] The task completion degree is evaluated based on the optimal cleaning strategy. The coverage range of the sweeper in the rasterized cleaning area is calculated according to the path plan to be evaluated. The number of effectively covered grids is counted. The multiple coverage areas and their time distributions are detected. The un-covered areas are identified using connected component analysis and their accessibility is evaluated. Supplementary cleaning paths are generated for the un-covered areas, and a task completion degree evaluation result is obtained based on the supplementary cleaning paths;

[0129] The resource consumption is evaluated based on the optimal cleaning strategy. The dynamic power consumption is calculated according to the path curvature data and the maximum allowable speed. The operation time is calculated based on the coverage range. The obstacle avoidance time is evaluated by combining the wheel-ground force model. The complexity of the path planning algorithm and the real-time control calculation amount are analyzed, and a resource consumption evaluation result is generated;

[0130] According to the execution feasibility evaluation result, the task completion degree evaluation result, and the resource consumption evaluation result, the optimal cleaning path is selected from the path plan to be evaluated; a secure transmission channel with end-to-end encryption function is constructed based on the selected optimal cleaning path.

[0131] First, determine the optimal cleaning strategy and perform regional rasterization and path plan generation. According to the characteristics of the cleaning environment (such as obstacle distribution, area shape, etc.) and the requirements of the cleaning task (such as coverage rate requirements, time limits, etc.), select a suitable cleaning strategy, such as bow-shaped cleaning, spiral cleaning, etc. Then, rasterize the cleaning area, discretize the continuous cleaning area into several grids of equal size. According to the selected cleaning strategy and the rasterization result, generate multiple path plans to be evaluated, and each path plan consists of a series of path points connecting the centers of adjacent grids. For example, in a 10m x 10m area, divided by 0.5m x 0.5m grids, multiple bow-shaped or spiral path plans can be generated.

[0132] Next, perform an execution feasibility assessment on the generated path plan. Perform Bezier curve fitting on each path plan to be evaluated to obtain a smooth path curve and path curvature data. Based on the path curvature data, establish a kinematic model to calculate the maximum allowable speed for each path segment. For example, at a curve with a large curvature, the maximum allowable speed will be reduced to ensure the stability of the cleaning robot. Then, based on the maximum allowable speed, use the zero moment point theory to analyze the attitude stability margin. For example, by analyzing the center of gravity offset and tilt angle of the cleaning robot at different speeds, determine whether it will tip over. Combine the attitude stability margin with ground characteristics (such as friction coefficient, ground slope, etc.) to establish a wheel-ground force model to verify the distance constraint between adjacent path points and speed continuity. For example, ensure that the distance between adjacent path points does not exceed the movement range of the cleaning robot, and the speed change is not too drastic. Finally, generate an execution feasibility assessment result, such as feasible / infeasible, or a scoring system.

[0133] Then, perform a task completion assessment on the path plan. According to each path plan to be evaluated, calculate the coverage range of the sweeper in the rasterized cleaning area and count the number of effectively covered grids. For example, count the number of grids completely covered by the sweeper and the number of partially covered grids. Detect the multiple coverage areas and their time distribution. For example, identify which areas are repeatedly cleaned and the number of times and time of repeated cleaning. Use connected component analysis to identify uncovered areas and evaluate their accessibility. For example, determine whether the uncovered areas are inaccessible due to obstacles or other reasons. For the uncovered areas, generate supplementary cleaning paths. For example, if there are accessible uncovered areas, generate a new path segment to clean them up. Based on the supplementary cleaning path, update the coverage range and obtain the task completion assessment result, such as the coverage percentage. Suppose a path plan covers 200 grids and the total number of grids is 250, then the coverage rate is 80%.

[0134] Subsequently, perform a resource consumption assessment on the path plan. Calculate the dynamic power consumption based on the path curvature data and the maximum allowable speed. For example, in the case of high-speed movement or frequent turning, the power consumption will be higher. Calculate the operation time based on the coverage range. For example, calculate the total time required to complete the cleaning task according to the cleaning speed and coverage area of the sweeper. Combine the wheel-ground force model to evaluate the obstacle avoidance time. For example, estimate the additional time required for obstacle avoidance based on the number of obstacles and the obstacle avoidance strategy. Analyze the complexity of the path planning algorithm and the real-time control calculation amount. For example, evaluate the computational efficiency and real-time performance of different path planning algorithms. Finally, generate a resource consumption assessment result, such as total power consumption, total time, calculation amount, etc. For example, the total power consumption of a certain path plan is 1000J and the total time is 30 minutes.

[0135] Finally, the optimal cleaning path is selected according to the evaluation results of each item. Based on the evaluation results of execution feasibility, task completion, and resource consumption, the optimal cleaning path is selected from the path schemes to be evaluated. For example, select a path scheme with high coverage, low power consumption, short time, and executability. Based on the selected optimal cleaning path, a secure transmission channel with end-to-end encryption function is constructed to securely transmit the path data to the cleaning robot.

[0136] The solution of this application can:

[0137] Improve cleaning efficiency: Through the optimal cleaning strategy and path planning algorithm, repeated cleaning and ineffective cleaning can be minimized to improve cleaning efficiency and shorten the cleaning time. Reduce resource consumption: By optimizing the path planning, the power consumption and loss of the cleaning robot can be reduced, its service life can be extended, and the energy cost can be lowered. Enhance security: By constructing a secure transmission channel with end-to-end encryption function, the security of path data can be effectively protected, preventing data leakage and tampering.

[0138] Figure 2 It is a schematic structural diagram of the progressive privacy computing system for the virtual-real fusion scenario of the intelligent sweeper in the embodiment of the present invention. As Figure 2 shown, the system includes:

[0139] The first unit is used to collect indoor space data through the multi-modal sensor array of the intelligent sweeper, obtain the point cloud data of the micrometer-level lidar, the three-dimensional image data of the depth perception camera, the spatial response data of the sonar sensor, and the reflection signal data of the millimeter-wave radar, and input the point cloud data, the three-dimensional image data, the spatial response data, and the reflection signal data into the entangled state registration network to generate a six-dimensional scene feature tensor in a unified coordinate system; input the six-dimensional scene feature tensor into the graph neural network for semantic segmentation and temporal modeling to construct a scene knowledge graph including the spatial distribution of scene objects, object attribute information, and spatio-temporal interaction relationships between objects;

[0140] The second unit is used to input the scene knowledge graph into the deep reinforcement learning network, construct a privacy sensitivity score matrix according to the object attribute information in the scene knowledge graph; based on the privacy sensitivity score matrix, use the zero-knowledge proof mechanism to perform hierarchical encryption processing on the nodes in the scene knowledge graph to generate a multi-level encrypted graph; submit the multi-level encrypted graph to the blockchain network node group, and calculate and output multiple scene data versions with different privacy protection intensities through the federated learning framework, and each scene data version includes corresponding privacy protection intensity parameters and scene information integrity parameters;

[0141] A third unit is configured to construct a path planning optimization model in a secure multi-party computing environment according to the scenario data version and its corresponding parameters; input the path planning optimization model into a neuro-symbolic inference engine to generate an initial set of cleaning path plans; use a graph attention network to perform distributed optimization evaluation on the initial set of cleaning path plans, and output an optimal cleaning strategy with a credibility quantification index; screen the optimal cleaning path based on the optimal cleaning strategy, and send the optimal cleaning path to the intelligent sweeper for execution through an encrypted channel; collect the real-time status data generated during the execution of the intelligent sweeper, and send the real-time status data back to the scenario knowledge graph for dynamic update to achieve continuous optimization of the scenario model.

[0142] In a third aspect of the embodiments of the present invention,

[0143] a kind of electronic device is provided, including:

[0144] a processor;

[0145] a memory for storing instructions executable by the processor;

[0146] wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0147] In a fourth aspect of the embodiments of the present invention,

[0148] a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0149] The present invention can be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are uploaded.

[0150] 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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A progressive privacy computing method for the virtual-real fusion scenario of an intelligent sweeping robot, characterized in that: include: The multimodal sensor array of the intelligent sweeper is used to collect indoor space data, obtain point cloud data of micron-level laser radar, three-dimensional image data of depth perception camera, spatial response data of sonar sensor and reflection signal data of millimeter-wave radar, and input the point cloud data, the three-dimensional image data, the spatial response data and the reflection signal data into the entangled state registration network to generate a six-dimensional scene feature tensor in a unified coordinate system; the six-dimensional scene feature tensor is input into the graph neural network for semantic segmentation and time series modeling, and a scene knowledge graph containing the spatial distribution of scene objects, object attribute information and the spatiotemporal interaction relationship between objects is constructed; Inputting the scene knowledge graph into a deep reinforcement learning network, and constructing a privacy sensitivity scoring matrix according to the object attribute information in the scene knowledge graph; Based on the privacy sensitivity scoring matrix, the nodes in the scenario knowledge graph are hierarchically encrypted using a zero-knowledge proof mechanism to generate a multi-level encrypted graph; the multi-level encrypted graph is submitted to the blockchain network node group, and multiple scenario data versions with different privacy protection strengths are calculated and output through a federated learning framework, each of which contains corresponding privacy protection strength parameters and scenario information integrity parameters; According to the scenario data version and its corresponding parameters, a path planning optimization model is constructed in a secure multi-party computing environment; the path planning optimization model is input into a neural symbolic reasoning engine to generate an initial cleaning path solution set; a graph attention network is used to perform distributed optimization evaluation on the initial cleaning path solution set, and an optimal cleaning strategy with a quantified trustworthiness index is output; an optimal cleaning path is selected based on the optimal cleaning strategy, and the optimal cleaning path is sent to the intelligent sweeper through an encrypted channel for execution; The real-time status data generated by the intelligent sweeper during execution is collected, and the real-time status data is transmitted back to the scene knowledge graph for dynamic updating to achieve continuous optimization of the scene model.

2. The method according to claim 1, characterized in that: Inputting the scene knowledge graph into a deep reinforcement learning network, and constructing a privacy sensitivity scoring matrix according to the object attribute information in the scene knowledge graph; Based on the privacy sensitivity scoring matrix, the nodes in the scenario knowledge graph are hierarchically encrypted using a zero-knowledge proof mechanism to generate a multi-level encrypted graph, including: The scene knowledge graph is input into a deep reinforcement learning network, wherein the deep reinforcement learning network comprises a state space module, an action space module and a value assessment module; the state space module receives spatial position information, object usage frequency information and interaction relationship information between objects in the scene knowledge graph; the action space module generates a privacy sensitivity score interval based on the output information of the state space module; the value assessment module calculates the privacy sensitivity score value according to the spatiotemporal correlation of object attributes and user privacy annotation information; The privacy sensitivity score value output by the deep reinforcement learning network is input into a scoring matrix construction module, and the scoring matrix construction module uses an adaptive weight fusion algorithm to normalize the privacy sensitivity score value to generate a privacy sensitivity scoring matrix that characterizes the degree of privacy association between object nodes in the scene knowledge graph; based on the privacy sensitivity scoring matrix, the object nodes are divided into sensitive protection levels higher than a preset sensitivity threshold and sensitive protection levels lower than the preset sensitivity threshold using a spectral clustering algorithm; A zero-knowledge proof mechanism is called to perform hierarchical encryption processing on object nodes with different sensitive protection levels, and the zero-knowledge proof mechanism includes a proof generation unit and a verification execution unit; the proof generation unit constructs an encryption proof using the Groth16 protocol, and adopts a fully homomorphic encryption scheme for object nodes with a sensitive protection level higher than a preset sensitivity threshold, and adopts a lightweight searchable encryption scheme for object nodes with a sensitive protection level lower than the preset sensitivity threshold; the verification execution unit verifies the legitimacy of the encrypted data through a pairing operation, outputs the encryption proof verification result, and reconstructs the encrypted object nodes that have passed the verification into a multi-level encryption graph.

3. The method according to claim 1, characterized in that The multi-level encrypted graph is submitted to the blockchain network node group, and multiple scene data versions with different privacy protection strengths are calculated and output through the federated learning framework. Each scene data version contains corresponding privacy protection strength parameters and scene information integrity parameters including: The multi-level encrypted graph is divided into data shards according to the spatial position relationship, and the data shards are distributed to multiple nodes in the blockchain network node group using a consistent hashing algorithm, wherein the data shards include node encryption attribute data, edge relationship encryption data, and shard index information; a practical Byzantine fault-tolerant consensus algorithm is used in the blockchain network node group to perform data consistency verification, calculation correctness verification, and version synchronization verification on the data shards; A federated learning framework is constructed based on the blockchain network node group, and the federated learning framework includes a feature extraction network, a feature fusion network and a decision network; the feature extraction network extracts features from encrypted data local to the node, the feature fusion network integrates feature information of multiple nodes, and the decision network generates scene data versions with different privacy protection strengths; an adaptive federated optimization algorithm is used in the federated learning framework for model training, and the adaptive federated optimization algorithm performs local model updates through nodes, reduces communication overhead through gradient compression, uses homomorphic encryption to aggregate model parameters, and introduces a differential privacy protection gradient update process; The output results of the federated learning framework are input into the scene data generation module, which generates privacy protection strength parameters according to the degree of data desensitization, encryption strength and noise level, and generates scene information integrity parameters according to topology preservation, attribute retention rate and relationship integrity.

4. The method according to claim 1, characterized in that: According to the scenario data version and its corresponding parameters, a path planning optimization model is constructed in a secure multi-party computing environment; the path planning optimization model is input into a neural symbolic reasoning engine to generate an initial cleaning path solution set; The graph attention network is used to perform distributed optimization evaluation on the initial cleaning path solution set, and the optimal cleaning strategy with credibility quantification indicators is output, including: According to the scene data version and its corresponding privacy protection strength parameter and scene information integrity parameter, a path planning optimization model is constructed in a secure multi-party computing environment; the path planning optimization model adopts a three-party computing protocol based on secret sharing to decompose the computing task into a spatial constraint computing task, a temporal constraint computing task and a result aggregation computing task; the path planning optimization model outputs an initial path planning solution that integrates physical constraints, environmental constraints and privacy constraints; The path planning initial solution is input into the neural symbolic reasoning engine for deep reasoning operation. The encoding layer of the neural symbolic reasoning engine converts the scene spatial relationship, action temporal relationship and constraint rule relationship in the path planning initial solution into a vector representation; the reasoning layer of the neural symbolic reasoning engine extracts implicit rules from historical data based on the vector representation and merges them with expert rules to generate a path planning rule set; the decoding layer of the neural symbolic reasoning engine restores the reasoning result to a symbolic path description based on the path planning rule set to generate an initial cleaning path solution set containing multiple candidate paths; Constructing a graph structure representation for the candidate paths in the initial cleaning path solution set, wherein nodes in the graph structure representation correspond to path points and edges correspond to path segments; inputting the graph structure representation into a graph attention network, and calculating node-level attention weights, edge-level attention weights, and global attention weights through distributed computing nodes; prioritizing the candidate paths based on the node-level attention weights, edge-level attention weights, and global attention weights, and generating a path solution sequence with evaluation scores; A credibility quantization index is calculated for each path in the path plan sequence, and the credibility quantization index is input into the graph attention network as evaluation feedback, the node-level attention weight, the edge-level attention weight and the global attention weight are updated, the cleaning path plan sequence is optimized and the optimal cleaning strategy is output.

5. The method according to claim 4, characterized in that Input the graph structure representation into the graph attention network, and calculate the node-level attention weight, edge-level attention weight and global attention weight through distributed computing nodes; Prioritizing candidate paths based on the node-level attention weight, the edge-level attention weight, and the global attention weight to generate a path solution sequence with evaluation scores includes: Acquire path features, wherein the path features include geometric features of path segment length, path segment curvature, and path segment slope, as well as performance features of energy consumption index and time index; construct the geometric features and the performance features into a node attribute vector, and construct the corner features and distance features between the path segments into an edge attribute vector; The graph structure is input into a graph attention network, and the node attribute vector and the edge attribute vector are processed in parallel by distributed computing nodes; the distributed computing nodes construct a query vector and a key vector based on the node attribute vector, and a self-attention mechanism is used to calculate the correlation between the query vector and the key vector to obtain a node attention weight; The distributed computing node concatenates the node attribute vector and the edge attribute vector and inputs them into a trainable parameter matrix, calculates the constraint relationship between adjacent path segments to obtain edge-level attention weights; the distributed computing node constructs a global context vector, concatenates the node attribute vector and the global context vector and inputs them into a multi-layer perceptron, calculates the correlation between the overall features and local features of the path to obtain a global attention weight; The distributed computing node performs a weighted combination of the node attention weight, the edge-level attention weight and the global attention weight based on preset weights to obtain a comprehensive score of the path; the candidate paths are prioritized according to the comprehensive score of the path to generate a path solution sequence with evaluation scores.

6. The method according to claim 1, characterized in that The optimal cleaning path is selected based on the optimal cleaning strategy, and the optimal cleaning path is sent to the intelligent sweeper through an encrypted channel for execution; the real-time status data generated by the intelligent sweeper during the execution process is collected, and the real-time status data is sent back to the scene knowledge graph for dynamic update, so as to realize continuous optimization of the scene model, including: Based on the optimal cleaning strategy, the path plan is evaluated for execution feasibility, task completion and resource consumption, and the optimal cleaning path is selected from the path plan according to the evaluation results; a secure transmission channel with end-to-end encryption function is constructed, and the secure transmission channel includes a hybrid encryption module, an identity authentication module and a dynamic key update module; the optimal cleaning path is sent to the intelligent sweeper through the secure transmission channel, wherein the sending process adopts segment confirmation and packet loss retransmission mechanism to ensure transmission reliability; Collect real-time status data from the intelligent sweeper, pre-process the real-time status data, perform noise filtering through Kalman filtering, achieve multi-source data fusion through time series alignment, perform anomaly detection through statistical analysis, and generate pre-processed status data; transmit the pre-processed status data back to the scene knowledge graph; The scene knowledge graph is dynamically updated through temporal knowledge fusion and spatial knowledge fusion. The temporal knowledge fusion is processed based on historical data weights, new data credibility and conflicting data reconciliation, and the spatial knowledge fusion is processed based on local update, global consistency and topology optimization. Based on the fusion results, the motion model parameters, environmental model parameters and task model parameters are optimized to achieve continuous optimization of the scene model.

7. The method according to claim 6, characterized in that Based on the optimal cleaning strategy, the path plan is evaluated for execution feasibility, task completion and resource consumption, and the optimal cleaning path is selected from the path plan according to the evaluation results; and a secure transmission channel with end-to-end encryption function is constructed, including: Based on the optimal cleaning strategy, the cleaning area is rasterized and divided, and multiple path plans to be evaluated are generated; based on the optimal cleaning strategy, an execution feasibility assessment is performed, Bezier curve fitting is performed on the path plan to be evaluated to obtain path curvature data, a kinematic model is established based on the path curvature data to calculate the maximum allowable speed of each path segment, the attitude stability margin is analyzed using the zero moment point theory according to the maximum allowable speed, a wheel-ground force model is established in combination with the attitude stability margin and ground features, the distance constraints of adjacent path points and the speed continuity are verified, and an execution feasibility assessment result is generated; Based on the optimal cleaning strategy, the task completion is evaluated, the coverage of the cleaner in the rasterized cleaning area is calculated according to the path plan to be evaluated, the number of effective coverage grids is counted, the multiple coverage areas and their time distribution are detected, the uncovered areas are identified by connected domain analysis and their accessibility is evaluated, and a supplementary cleaning path is generated for the uncovered areas to obtain the task completion evaluation result; Perform resource consumption evaluation based on the optimal cleaning strategy, calculate dynamic power consumption according to the path curvature data and the maximum allowable speed, calculate operation time based on the coverage range, evaluate obstacle avoidance time in combination with the wheel-ground force model, analyze the complexity of the path planning algorithm and the amount of real-time control calculation, and generate a resource consumption evaluation result; According to the execution feasibility assessment result, the task completion assessment result and the resource consumption assessment result, an optimal cleaning path is selected from the path schemes to be evaluated; and a secure transmission channel with end-to-end encryption function is constructed based on the optimal cleaning path.

8. A progressive privacy computing system for a virtual-reality fusion scene of an intelligent sweeping robot, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to collect indoor space data through the multimodal sensor array of the intelligent sweeper, obtain point cloud data of the micron-level laser radar, three-dimensional image data of the depth perception camera, spatial response data of the sonar sensor and reflection signal data of the millimeter wave radar, and input the point cloud data, the three-dimensional image data, the spatial response data and the reflection signal data into the entangled state registration network to generate a six-dimensional scene feature tensor in a unified coordinate system; input the six-dimensional scene feature tensor into the graph neural network for semantic segmentation and time series modeling, and construct a scene knowledge graph containing the spatial distribution of scene objects, object attribute information and the spatiotemporal interaction relationship between objects; The second unit is used to input the scene knowledge graph into a deep reinforcement learning network, and construct a privacy sensitivity scoring matrix according to the object attribute information in the scene knowledge graph; Based on the privacy sensitivity scoring matrix, the nodes in the scenario knowledge graph are hierarchically encrypted using a zero-knowledge proof mechanism to generate a multi-level encrypted graph; the multi-level encrypted graph is submitted to the blockchain network node group, and multiple scenario data versions with different privacy protection strengths are calculated and output through a federated learning framework, each of which contains corresponding privacy protection strength parameters and scenario information integrity parameters; The third unit is used to build a path planning optimization model in a secure multi-party computing environment according to the scenario data version and its corresponding parameters; input the path planning optimization model into a neural symbolic reasoning engine to generate an initial cleaning path solution set; use a graph attention network to perform distributed optimization evaluation on the initial cleaning path solution set, and output an optimal cleaning strategy with a trustworthy quantitative indicator; select an optimal cleaning path based on the optimal cleaning strategy, and send the optimal cleaning path to the intelligent sweeper through an encrypted channel for execution; The real-time status data generated by the intelligent sweeper during execution is collected, and the real-time status data is transmitted back to the scene knowledge graph for dynamic updating to achieve continuous optimization of the scene model.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images