A high-voltage live working intelligent robot system based on four-dimensional fusion and autonomous cooperation and a method thereof

Abstract

The application provides a high-voltage live-line work intelligent robot system and method based on four-dimensional fusion and autonomous cooperation, and belongs to the technical field of high-voltage live-line work safety management and control. The system comprises a four-dimensional data sensing module, a robot autonomous cooperation unit, a composite risk assessment and decision module, and an execution interaction module. The four-dimensional data sensing module is connected with the robot autonomous cooperation unit, the composite risk assessment and decision module is connected with the robot autonomous cooperation unit, and the execution interaction module is connected with the composite risk assessment and decision module. Through four-dimensional data fusion of environment, action, physiology and equipment, accurate identification of composite risks is realized, the risk identification coverage is increased from less than or equal to 30% to greater than or equal to 92%, the false alarm rate is reduced from greater than or equal to 40% to less than or equal to 10%, the robot has autonomous movement, tool carrying and automatic switching functions, the personnel load and the frequency of going up and down the tower are reduced, and the working time is expected to be shortened by greater than or equal to 40%.

Description

Technical Field

[0001] This invention relates to the field of safety management and control technology for live-line work, and in particular to an intelligent robot system and method for live-line work based on four-dimensional fusion and autonomous collaboration. Background Technology

[0002] Live-line working with high voltage is a high-risk aspect of equipment operation and maintenance in industries such as power, chemical, and mining. Existing technologies have explored ways to improve operational safety primarily in the following areas: High-voltage operation robots: Existing robots mostly have a single function (such as conductor inspection), rely on preset programs, lack real-time perception of dynamic environment and the ability to coordinate with personnel, and are not specially designed for the complex structure of high-voltage towers.

[0003] Work safety monitoring system: Common systems are usually based on visual recognition of safety equipment wearing. The monitoring dimensions are single and they fail to link with personnel physiological status and equipment operating status. They cannot identify complex safety hazards caused by personnel fatigue, equipment abnormalities and environmental risks.

[0004] Tool and equipment management system: Existing devices mostly focus on the registration and management of issuance and return, lacking real-time online detection of core safety indicators for high-voltage operations and insulation performance, and cannot prevent the use of defective tools with insulation failure.

[0005] Multidimensional monitoring technology: Various environmental monitoring devices (temperature, humidity, wind speed, etc.) typically issue alarms independently, lacking data fusion. This leads to a disconnect between early warnings and specific operational procedures and intervention measures, resulting in high false alarm rates that are easily overlooked by workers. Therefore, it is necessary to design an intelligent robot system and method for high-voltage live-line working based on four-dimensional fusion and autonomous collaboration. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent robot system and method for high-voltage live-line work based on four-dimensional fusion and autonomous collaboration, which solves the technical problems of existing high-voltage live-line work monitoring technology, such as incomplete risk monitoring dimensions, lack of human-machine collaboration, extensive tool management, disconnect between risk decision-making and proactive intervention, and inability to form an optimized closed loop of work data.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A high-voltage live-line working intelligent robot system based on four-dimensional fusion and autonomous collaboration includes a four-dimensional data perception module, a robot autonomous collaboration unit, a composite risk assessment and decision-making module, and an execution interaction module. The four-dimensional data perception module is connected to the robot autonomous collaboration unit, the composite risk assessment and decision-making module is connected to the robot autonomous collaboration unit, and the execution interaction module is connected to the composite risk assessment and decision-making module. The four-dimensional data perception module is used to collect several modal data of the high-voltage live-line working site in real time. The robot autonomous collaboration unit is used to assist humans in effectively performing various work tasks in the high-voltage live-line environment. The execution interaction module is used to display the operations to be performed on the robot autonomous collaboration unit and notify the user in real time. The composite risk assessment and decision-making module is used to analyze the fused multi-dimensional data, transform the complex site conditions into quantified risk levels, and automatically trigger precise intervention commands, thereby realizing the leap from passive alarm to active prevention and control.

[0008] Furthermore, the four-dimensional data perception module includes an environmental safety perception submodule, a personnel movement perception submodule, a physiological state perception submodule, and an equipment status perception submodule. The environmental safety perception submodule uses high-definition cameras and LiDAR to identify the laying of insulating mats, unauthorized personnel intrusion, and to measure the safe distance of machinery. The personnel movement perception submodule uses millimeter-wave sensors and skeletal point extraction algorithms to capture and analyze whether the body posture and operation trajectory of the workers comply with safety regulations. The physiological state perception submodule collects heart rate and body temperature data in real time through smart wearable devices worn by workers to determine whether there is excessive fatigue or sudden physical discomfort. The equipment status perception submodule integrates an infrared thermal imager and a partial discharge detector to monitor abnormal temperatures and partial discharge phenomena in cable joints and the equipment body.

[0009] Furthermore, the robot's autonomous collaborative unit includes a composite climbing mechanism, safety protection devices, an intelligent tool and material compartment, and an auxiliary operation module. The composite climbing mechanism adopts a wheel-type adsorption and gripper-type composite structure, enabling it to adapt to circular or square high-voltage towers for climbing and positioning. The safety protection device has a built-in electromagnetic anti-fall self-locking mechanism that can quickly lock in the event of a power outage or malfunction, ensuring the safety of the robot itself. The modular tool library in the intelligent tool and material compartment can carry several types of working tools and has a built-in megohmmeter that can detect the insulation resistance of tools in real time; if the insulation fails to meet the requirements, it will automatically lock. It is equipped with a storage compartment to deliver materials to personnel working at heights. The auxiliary operation module integrates LED lighting equipment to meet the needs of nighttime operations. Furthermore, the composite risk assessment and decision-making module includes a data preprocessing module, a feature extraction module, a dynamic weight optimization module, a risk quantification and classification module, and a graded intervention decision-making module. The data preprocessing module performs outlier processing, standardization, and temporal alignment on the four-dimensional perception data. The feature extraction module uses principal component analysis, bidirectional long short-term memory networks, and graph neural networks to extract static features, temporal features, and human-machine-environment correlation features of the data, respectively. The dynamic weight optimization module sets the initial weights for each dimension based on the analytic hierarchy process and introduces a deep Q-network reinforcement learning algorithm to dynamically adjust the weights based on feedback from actual operational data, enabling the assessment model to adapt to different scenarios. The risk quantification and classification module uses an ensemble learning model composed of random forests, support vector machines, and backpropagation neural networks to fuse and analyze the extracted features, output a quantified risk score, and classify it into risk levels of 0-4 based on historical data clustering. The graded intervention decision-making module automatically generates and executes corresponding intervention instructions based on the risk level, and provides early warning or process control through robots and personnel intelligent terminals.

[0010] Furthermore, the robot's autonomous collaborative unit also includes a safety protection mechanism. The safety protection mechanism has a built-in electromagnetic anti-fall self-locking device, which can quickly lock within 0.5 seconds in the event of power failure or malfunction, and has a maximum load capacity of ≥200kg, ensuring safety during operation.

[0011] Furthermore, the robot's autonomous collaborative unit also includes a positioning system that integrates GPS, LiDAR, and visual SLAM technologies, achieving a positioning accuracy of ±1cm, meeting the stringent requirements for precise operation in high-pressure work.

[0012] A method for a high-voltage live-line working intelligent robot system based on four-dimensional fusion and autonomous collaboration, the method comprising the following steps: Step 1: Before the operation, the robot adapts its climbing mode according to the tower structure, automatically checks the insulation performance of the tools, and associates the work order; Step 2: The robot autonomously climbs to the work site, and the four-dimensional data perception module is activated to collect environmental, personnel, and equipment data in real time; Step 3: The composite risk assessment and decision-making module processes the fused data and calculates the real-time risk level; Step 4: Based on the risk level, the system executes corresponding intervention measures, while the robot provides collaborative operation support such as tool switching and material replenishment according to instructions; Step 5: After the task is completed, the robot returns, and the data closed-loop management module archives all data from this task and generates an analysis report.

[0013] The present invention, by adopting the above-described technical solution, has the following beneficial effects: (1) This invention achieves accurate identification of complex risks through the fusion of four-dimensional data of environment, action, physiology and equipment, increasing the risk identification coverage from ≤30% to ≥92% and the false alarm rate from ≥40% to ≤10%. The robot has autonomous movement, tool carrying and automatic switching functions, reducing the load on personnel and the frequency of going up and down the tower, and is expected to shorten the operation time by ≥40%. It realizes online real-time detection of tool insulation performance and work order linkage verification, and prevents tools with insulation failure from being put on the job. The robot structure is specially adapted to high-voltage towers and has fall protection and night lighting functions. It can be widely used in various high-altitude and high-risk operation scenarios such as power and chemical industries.

[0014] (2) Closed-loop tool management reduces the rate of tools with insulation failure from ≥25% to ≤0.5%, improves tool traceability efficiency by ≥90%, and strictly meets the safety requirements of high-voltage operations; strong adaptability: compatible with round / square towers and day / night operation scenarios, it can adapt to the high-voltage operation needs of different industries without reconstructing the architecture, reducing adaptation costs by ≥80%; maximizes data value: operation data is traceable and analyzable, promoting the transformation of safety management from experience-driven to data-driven, and providing support for long-term safety management optimization. Attached Figure Description

[0015] Figure 1 This is a system module block diagram of the present invention; Figure 2 This is a top view of the cement kiln smoke chamber of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and preferred embodiments. However, it should be noted that many details listed in the specification are merely to provide the reader with a thorough understanding of one or more aspects of the present invention, and these aspects of the invention can be implemented even without these specific details.

[0017] like Figure 1-2 As shown, a high-voltage live-line working intelligent robot system based on four-dimensional fusion and autonomous collaboration includes a four-dimensional data perception module, a robot autonomous collaboration unit, a composite risk assessment and decision-making module, and an execution interaction module. The four-dimensional data perception module is connected to the robot autonomous collaboration unit, and the composite risk assessment and decision-making module is connected to the robot autonomous collaboration unit. The system adopts a four-level modular architecture: perception layer, data layer, decision layer, and execution layer. Each layer communicates via industrial Ethernet (EtherNet / IP) and 5G dual-mode to ensure low latency (≤500ms) and high reliability of data transmission. The perception layer includes a multimodal sensor cluster (environment, motion, physiology, and equipment) and robot body sensors; the data layer consists of edge computing nodes (local data preprocessing) and a cloud database (structured / time-series data storage); the decision layer integrates a composite risk assessment model and a graded intervention instruction generation module; and the execution layer encompasses the robot's autonomous operation unit, early warning interaction unit, and process locking unit.

[0018] 1) Four-dimensional data perception module Environmental safety perception: It adopts a 1080P high-definition smart camera and LiDAR. The camera has a frame rate of 30fps, a field of view of 120°, an IP66 waterproof rating, and an operating temperature range of -20℃ to 60℃. The LiDAR has a ranging range of 0.5-100m and an accuracy of ±2cm. It can realize the identification of insulation mat laying, the detection of unauthorized personnel intrusion, and the judgment of mechanical safety distance.

[0019] Personnel motion perception: Equipped with a millimeter-wave sensor and a skeletal point extraction module. The millimeter-wave sensor has a detection range of 0.1-10m and a motion capture frame rate of 15fps. The skeletal point extraction module can identify 33 key skeletal points with a recognition latency of ≤300ms, and can determine whether the worker's body tilt angle and tool operation trajectory comply with safety regulations.

[0020] Physiological state perception: Data is collected through wearable smart terminals (watches / goggles). The watch can monitor heart rate in the range of 60-200 beats / min with an accuracy of ±1 beat / min, and body temperature in the range of 35-42℃ with an accuracy of ±0.1℃. It can detect excessive fatigue or sudden discomfort in people in real time.

[0021] Equipment status awareness: Integrates an infrared thermal imager and a partial discharge detector. The infrared thermal imager has a temperature measurement range of -20℃ to 300℃ and an accuracy of ±1℃. The partial discharge detector has a detection sensitivity of ≤5pC and can accurately identify potential hazards such as overheating of cable joints and partial discharge of equipment.

[0022] (2) Robot Autonomous Collaboration Unit Climbing mechanism: It adopts a "wheel adsorption + gripper composite structure". The wheel adsorption is suitable for round towers with a diameter of 300-800mm, and the gripper is suitable for square towers with a side length of 200-500mm. The climbing speed is 0.1-0.3m / s, which can flexibly adapt to different types of tower structures.

[0023] Positioning system: Integrating GPS, LiDAR and visual SLAM technologies, the positioning accuracy reaches ±1cm, meeting the stringent requirements of precise operation in high-pressure work.

[0024] Safety protection: Built-in electromagnetic anti-fall self-locking device, which can lock quickly within 0.5 seconds in case of power failure or malfunction, with a maximum load capacity of ≥200kg, ensuring safety during operation.

[0025] Tool and Material Management: The modular tool library can hold 5-8 types of high-voltage working tools (such as wrenches, testers, etc.), with a tool switching time of ≤3s; the tool library has a built-in megohmmeter module to monitor the insulation resistance of tools in real time, with a threshold set at ≥100MΩ. When the resistance is lower than the threshold, the tool is automatically locked and an alarm is triggered; it is equipped with a 10L sealed storage compartment, which can carry tools, food, drinking water and other supplies, and deliver them to the workers precisely via wireless control.

[0026] Nighttime adaptation: Integrated with a 30W LED floodlight, with an illumination distance of ≥20m, a color temperature of 5000K, and supports automatic light-sensing activation function to meet the lighting needs of nighttime operations.

[0027] (3) Composite Risk Assessment and Decision-Making Module The core of this module is a five-level algorithm chain: "data preprocessing - feature extraction - dynamic weight optimization - multi-model fusion risk quantification - decision instruction generation", which is implemented as follows: Step 1: Data Preprocessing Algorithm Outlier handling: An improved 3σ criterion combined with the Isolation Forest algorithm is used to identify and correct outliers (such as sensor false alarms and data mutations) in the four-dimensional sensing data. The number of decision trees in the Isolation Forest is set to 100, and the anomaly detection threshold is set to 0.6 to ensure data reliability.

[0028] Data standardization: Z-score standardization is applied to data of different dimensions (such as heart rate, temperature, distance, and insulation resistance) to map the data to the [0,1] interval. The formula is: x'=(x-μ) / σ (where μ is the mean and σ is the standard deviation), which eliminates the influence of dimensional differences on the evaluation results.

[0029] Timing data alignment: Based on the "timestamp interpolation method", data with different sampling frequencies (physiological data 1Hz, device data 0.5Hz, action data 15fps) are uniformly aligned to a sampling frequency of 1Hz. Linear interpolation is used to supplement missing data to ensure timing consistency.

[0030] Step 2: Multidimensional Feature Extraction Algorithm Static feature extraction: Principal component analysis (PCA) is used to reduce the dimensionality of the standardized four-dimensional data, retaining principal components with a cumulative variance contribution rate of ≥95% (3-5 principal components are expected to be retained), reducing data redundancy and improving the computational efficiency of the algorithm.

[0031] Temporal feature extraction: Based on the temporal correlation of physiological data and device status data, a bidirectional long short-term memory network (Bi-LSTM) is used to extract dynamic features. The network structure is set to 2 hidden layers with 64 neurons in each layer, a dropout rate of 0.2, and the activation function is ReLU to capture the trend changes of data in the past 5 minutes (such as continuous increase in heart rate and gradual increase in device temperature exceeding the standard).

[0032] Feature extraction: A graph neural network (GNN) is used to construct a "personnel-equipment-environment" association graph. Four-dimensional data is used as node features. The association strength between nodes is defined by the adjacency matrix (e.g., the association weight of "equipment high temperature" and "person approaching action" is set to 0.8). Composite risk association features (e.g., the combined feature of "equipment high temperature + person approaching illegally + heart rate increase") are extracted.

[0033] Step 3: Dynamic Weight Optimization Algorithm Initial weight setting: Based on the analytic hierarchy process (AHP), five high-pressure operation safety experts were invited to score the risk impact of the four-dimensional data, construct a judgment matrix and calculate the initial weights: basic weight of operation type (0.3), action compliance weight (0.25), physiological state weight (0.25), equipment state weight (0.15), and environmental coefficient weight (0.05).

[0034] Dynamic weight adjustment: A Deep Q-Network (DQN) reinforcement learning algorithm is introduced, with the reward function being "maximizing risk identification accuracy and minimizing false alarm rate," to adjust the weights of each dimension in real time. The reward function is designed as: R = α × Acc - β × FalseAlarm (where α = 0.7, β = 0.3, Acc is the identification accuracy, and FalseAlarm is the false alarm rate). The DQN network iteration count is set to 1000 times / day, and the learning rate is set to 0.001. The weights are adaptively optimized based on daily task data to cope with dynamic changes in the scenario (such as increasing the weight of the environmental coefficient in high-temperature environments and increasing the weight of the physiological state when personnel are experiencing physiological abnormalities).

[0035] Step 4: Multi-model fusion risk quantification algorithm Base model selection: Random forest (RF), support vector machine (SVM), and backpropagation neural network are used as base models to build an ensemble learning framework.

[0036] Random Forest: The number of decision trees is set to 200, and the maximum tree depth is set to 10, in order to capture non-linear relationships in the data; Support Vector Machine: The kernel function is RBF kernel, the penalty coefficient C is set to 10, and the gamma value is set to 0.1 to improve the generalization ability in small sample scenarios; Backpropagation (BP) neural network: The input layer is the extracted feature vector (dimensions 3-5), there is 1 hidden layer (32 neurons), and the output layer is the risk score (0-150 points), which is used to fit complex mapping relationships.

[0037] Model fusion strategy: The weighted voting method is used to fuse the output results of the base models. The weights are dynamically allocated according to the real-time accuracy of each model (e.g., if RF accuracy is 92%, SVM accuracy is 88%, and BP accuracy is 90%, then the weights are 0.35, 0.3, and 0.35 respectively). The final risk score = RF score × 0.35 + SVM score × 0.3 + BP score × 0.35.

[0038] Risk level classification: Based on the K-means clustering algorithm, historical risk scores are clustered to determine the optimal classification thresholds: 0-49 points (Level 0, safe), 50-69 points (Level 1, warning), 70-89 points (Level 2, alarm), 90-109 points (Level 3, locking non-critical steps), ≥110 points (Level 4, emergency shutdown + rescue linkage).

[0039] Step 5: Decision Instruction Generation and Optimization Command matching: A command library is built based on risk level and operation scenario (inspection / maintenance / emergency repair). A rule engine (Drools) is used to achieve accurate matching between risk level and intervention command. The rule library supports visual configuration and dynamic updates.

[0040] Command priority sorting: When there are multiple risks superimposed (such as high equipment temperature + personnel physiological abnormality), a greedy algorithm is used to sort the command priorities, and the intervention commands corresponding to high risks are executed first (such as level 4 commands have higher priority than level 3 commands, and equipment safety-related commands have higher priority than fatigue warning commands).

[0041] Instruction feedback optimization: The execution results (such as whether the warning is effective or whether the process locking is reasonable) are used as feedback signals and input to the dynamic weight optimization module to continuously optimize the algorithm parameters and instruction matching rules.

[0042] Step 6: Algorithm Adaptive Learning Offline training: The algorithm model is updated offline every day at midnight using the previous day's task data (including four-dimensional perception data, risk events, and intervention results). Incremental training is used (only the model parameters corresponding to newly added data are updated) to avoid repeated training. The training time is ≤1 hour.

[0043] Online inference optimization: Real-time calculation of model prediction accuracy and false alarm rate. When the accuracy is below 85% or the false alarm rate is above 12% for 10 consecutive minutes, online fine-tuning is triggered to dynamically adjust the number of hidden layer neurons in Bi-LSTM or the number of decision trees in RF to ensure that the algorithm adapts to changes in the scenario.

[0044] Data fusion logic: Based on the three-dimensional key fields of "timestamp - personnel ID - job ID", the preprocessed four-dimensional data is associated with the extracted feature vectors to form a unified data chain. For example, "2024-05-20 10:30 - personnel A - line inspection" can be associated with "standardized ambient temperature 0.6 (corresponding to 35℃) + action compliance 0.9 (corresponding to tilt 15°) + physiological state 0.7 (corresponding to heart rate 120 beats / min) + equipment status 0.5 (corresponding to temperature 85℃) + time-series feature vector [0.62, 0.58, 0.71] + associated feature 0.8 (association between high equipment temperature and personnel action)", providing comprehensive support for algorithm calculation.

[0045] (4) Data closed-loop management module Storage architecture: Local edge nodes can cache data for 72 hours and support the function of resuming transmission after network interruption to avoid data loss; the cloud adopts a "PostgreSQL + InfluxDB" combined storage solution with a data retention period of ≥3 years to meet long-term traceability needs.

[0046] Analysis and Application: The system can automatically generate work reports, including core content such as statistics on violations, distribution of equipment hazards, and physiological state change curves, providing data-driven support for the development of safety training plans and the optimization of work processes.

[0047] This invention is not simply a combination of "robot + sensor + algorithm," but rather achieves a creative breakthrough through the following collaborative design: Firstly, it integrates "physiological state perception" into a high-voltage live-line working robot system, constructing a four-dimensional fusion mechanism of "environment-action-physiology-equipment," fundamentally solving the pain point of existing technologies' inability to identify complex risks. Secondly, it specifically designs a tool management logic of "insulation detection + work order linkage + full lifecycle traceability," accurately addressing the core safety hazard of "tool insulation failure" in high-voltage scenarios, differing from simple requisition registration in ordinary scenarios. Thirdly, it establishes a strong correlation mechanism of "risk quantification scoring - process hierarchical locking," achieving a leap from "passive early warning" to "active prevention," overcoming the limitations of existing technologies' independent alarms. Fourthly, the robot's climbing mechanism is precisely adapted to the high-voltage tower structure, integrating scenario-based functions such as fall protection, nighttime lighting, and material replenishment; it is not a simple modification of a general-purpose robot, but fully adapts to the special needs of high-voltage operations.

[0048] Significantly Reduced Safety Risks: The coverage rate of composite risk identification has increased from ≤30% in existing technologies to ≥92%, the false alarm rate has decreased from ≥40% to ≤10%, and the incidence of high-voltage operation accidents is expected to decrease by ≥70%, greatly improving operational safety. Increased Operational Efficiency: The robot's autonomous movement and automatic tool changing capabilities shorten operation time by ≥40%; tool carrying capabilities reduce personnel load by ≥50%, reducing operational errors caused by physical exertion. Closed-Loop Tool Management: The rate of tools with insulation failures being used has decreased from ≥25% to ≤0.5%, and tool traceability efficiency has increased by ≥90%, strictly meeting the safety requirements of high-voltage operations. Strong Adaptability: Compatible with round / square towers and day / night operation scenarios, it can adapt to the high-voltage operation needs of different industries without refactoring the architecture, reducing adaptation costs by ≥80%. Maximized Data Value: Operational data is traceable and analyzable, driving the transformation of safety management from "experience-driven" to "data-driven," providing support for long-term safety management optimization.

[0049] Taking "110kV high-voltage line inspection operation" as an example, the implementation process is explained in detail: 1. Task Preparation Stage Robot deployment: Depending on the tower type (circular, 500mm in diameter), switch to wheel-type adsorption climbing mode and install inspection tools (infrared detector, ultrasonic flaw detector). Tool verification: The robot automatically detects the insulation resistance of the tools (all ≥150MΩ), links it to the work order (number: XJ2024052001), and confirms that the tools match the work requirements; Personnel adaptation: Operators wear smartwatches and goggles, the system binds personnel ID and robot ID, and initializes physiological baseline data (heart rate 75 beats / min, body temperature 36.5℃).

[0050] 2. Operation Execution Phase Autonomous movement: The robot climbs along the tower to the working position (15m high), locates itself using GPS + LiDAR (accuracy ±1cm), and activates the fall protection self-locking mechanism; Multi-dimensional monitoring: Cameras confirmed that the insulating mats were properly laid; LiDAR confirmed that no unauthorized personnel had entered; millimeter-wave sensors monitored personnel's operational trajectories (compliant with safety regulations); infrared thermal imagers detected a cable joint temperature of 98℃ (exceeding the standard). Risk assessment: System-related data (base score 60 points + action correction score + 5 points + physiological correction score 0 points + equipment correction score - 20 points) × environmental coefficient 1.2 (temperature 35℃) = 54 points (Level 1 risk); Intervention Execution: The goggles display a text message: "Cable joint temperature exceeds the limit, please pay attention to safety." The robot then pauses its approach to the area and waits for personnel to handle the situation.

[0051] Tool switching and material replenishment: During operation, the robot automatically switches tools according to needs. If the insulation performance of a tool is abnormal, an alarm is immediately triggered and a spare tool is replaced. When the operation time is too long, the robot delivers food and other supplies.

[0052] 3. End of assignment phase The robot autonomously returns to the ground carrying the tools, automatically returns the tools, and retests the insulation performance. The cloud-based operation report records one potential equipment hazard (overheating of a cable connector), no abnormalities in the personnel's physiological condition, and an operation duration of 2.5 hours (1.5 hours shorter than traditional operations). Data application: The report is synchronized to the safety management department for subsequent equipment maintenance and work process optimization.

[0053] Matters not covered in this invention are common knowledge.

[0054] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A high-voltage live-line working intelligent robot system based on four-dimensional fusion and autonomous collaboration, characterized in that: It includes a four-dimensional data perception module, a robot autonomous collaboration unit, a composite risk assessment and decision-making module, and an execution interaction module. The four-dimensional data perception module is connected to the robot autonomous collaboration unit, the composite risk assessment and decision-making module is connected to the robot autonomous collaboration unit, and the execution interaction module is connected to the composite risk assessment and decision-making module. The four-dimensional data perception module is used to collect several modal data of the high-voltage live-line working site in real time. The robot autonomous collaboration unit is used to assist humans in performing various work tasks in the high-voltage live-line environment. The execution interaction module is used to display the operations to be performed on the robot autonomous collaboration unit and notify the user in real time. The composite risk assessment and decision-making module is used to analyze the fused multi-dimensional data, transform the complex site conditions into quantified risk levels, and automatically trigger precise intervention commands, thereby realizing the leap from passive alarm to active prevention and control.

2. The intelligent robot system for high-voltage live-line working based on four-dimensional fusion and autonomous collaboration as described in claim 1, characterized in that: The four-dimensional data perception module includes an environmental safety perception submodule, a personnel movement perception submodule, a physiological state perception submodule, and an equipment status perception submodule. The environmental safety perception submodule uses high-definition cameras and LiDAR to identify the laying of insulating mats, unauthorized personnel intrusion, and to measure the safe distance of machinery. The personnel movement perception submodule uses millimeter-wave sensors and skeletal point extraction algorithms to capture and analyze whether the body posture and operation trajectory of the workers comply with safety regulations. The physiological state perception submodule collects heart rate and body temperature data in real time through smart wearable devices worn by workers to determine whether there is excessive fatigue or sudden physical discomfort. The equipment status perception submodule integrates an infrared thermal imager and a partial discharge detector to monitor abnormal temperatures and partial discharge phenomena in cable joints and equipment bodies.

3. The intelligent robot system for high-voltage live-line working based on four-dimensional fusion and autonomous collaboration as described in claim 1, characterized in that: The robot's autonomous collaborative unit includes a composite climbing mechanism, safety protection devices, an intelligent tool and material compartment, and an auxiliary operation module. The composite climbing mechanism adopts a wheel-type adsorption and gripper-type composite structure, enabling it to adapt to circular or square high-voltage towers for climbing and positioning. The safety protection device has a built-in electromagnetic anti-fall self-locking mechanism that can quickly lock in the event of power failure or malfunction, ensuring the safety of the robot itself. The modular tool library in the intelligent tool and material compartment can carry several types of working tools and has a built-in megohmmeter that can detect the insulation resistance of tools in real time; if the insulation fails to meet the standard, it will automatically lock. It is equipped with a storage compartment to deliver materials to personnel working at heights. The auxiliary operation module integrates LED lighting equipment to meet the needs of nighttime operations.

4. The intelligent robot system for live-line working of high voltage based on four-dimensional fusion and autonomous collaboration as described in claim 1, characterized in that: The composite risk assessment and decision-making module includes a data preprocessing module, a feature extraction module, a dynamic weight optimization module, a risk quantification and classification module, and a graded intervention decision-making module. The data preprocessing module performs outlier processing, standardization, and temporal alignment on the four-dimensional perception data. The feature extraction module uses principal component analysis, bidirectional long short-term memory networks, and graph neural networks to extract static features, temporal features, and human-machine-environment correlation features of the data, respectively. The dynamic weight optimization module sets the initial weights for each dimension based on the analytic hierarchy process and introduces a deep Q-network reinforcement learning algorithm to dynamically adjust the weights based on feedback from actual operational data, making the assessment model adaptable to different scenarios. The risk quantification and classification module uses an ensemble learning model composed of random forests, support vector machines, and backpropagation neural networks to fuse and analyze the extracted features, output a quantified risk score, and classify it into risk levels of 0-4 based on historical data clustering. The graded intervention decision-making module automatically generates and executes corresponding intervention instructions based on the risk level, and provides early warning or process control through robots and personnel intelligent terminals.

5. The intelligent robot system for high-voltage live-line working based on four-dimensional fusion and autonomous collaboration according to claim 3, characterized in that: The robot's autonomous collaborative unit also includes a safety protection mechanism. The safety protection mechanism has a built-in electromagnetic anti-fall self-locking device, which can lock quickly within 0.5 seconds in the event of power failure or malfunction, and has a maximum load capacity of ≥200kg, ensuring safety during operation.

6. The intelligent robot system for high-voltage live-line working based on four-dimensional fusion and autonomous collaboration according to claim 3, characterized in that: The robot's autonomous collaborative unit also includes a positioning system that integrates GPS, LiDAR, and visual SLAM technologies, achieving a positioning accuracy of ±1cm, meeting the stringent requirements for precise operation in high-pressure work.

7. A method for a high-voltage live-line working intelligent robot system based on four-dimensional fusion and autonomous collaboration according to any one of claims 1-6, characterized in that, The method includes the following steps: Step 1: Before the operation, the robot adapts its climbing mode according to the tower structure, automatically checks the insulation performance of the tools, and associates the work order; Step 2: The robot autonomously climbs to the work site, and the four-dimensional data perception module is activated to collect environmental, personnel, and equipment data in real time; Step 3: The composite risk assessment and decision-making module processes the fused data and calculates the real-time risk level; Step 4: Based on the risk level, the system executes corresponding intervention measures, while the robot provides collaborative operation support such as tool switching and material replenishment according to instructions; Step 5: After the task is completed, the robot returns, and the data closed-loop management module archives all data from this task and generates an analysis report.

Images