Flying robot inspection system and method based on Internet of Things

Abstract

The invention discloses a flying robot inspection system and method based on the Internet of Things, and relates to the technical field of big data analysis. Complete flight data of multiple devices of the same type and the same specification are screened, each inspection state characteristic parameter and the corresponding power consumption value are analyzed and separated in a structured mode, and a double power consumption value prediction mechanism is constructed; according to the method, data subsets are split by a flight state characteristic parameter factory, an exclusive service interface is configured, a first inspection power consumption value is output through a neural network model based on historical common data training, and a second inspection power consumption value is obtained through cooperative training by combining real-time parameters and recent historical data intercepted by a height return flight sliding window; the minimum inspection power consumption value needed by return flight is accurately calculated by constructing the return flight height mapping table, the return flight request is generated through comparison of the remaining electric quantity, the risks of contact loss and falling damage caused by insufficient electric quantity of the flying and climbing robot are effectively avoided, the safety and reliability of inspection operation are greatly improved, and stable application of equipment in complex scenes is guaranteed.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of big data analysis, and particularly relates to a flying robot inspection system and method based on Internet of Things. BACKGROUND

[0002] The flying robot can adapt to high-altitude facilities, narrow channels, irregular structures and other multi-element inspection scenes, can break through the operation limitations of traditional single motion mode equipment, can meet the core needs of precise inspection in complex environments in the fields of industrial production, infrastructure operation and maintenance, security patrol and the like, and can provide key technical support for efficient implementation of various inspection tasks, and the technical application has irreplaceable important significance for improving the coverage range and implementation feasibility of the inspection operation.

[0003] The flying robot is an inspection equipment with flying and crawling compound motion capability, and the power consumption rate of the flying robot in the shape switching process of crawling to flying and flying to crawling and in different motion states such as ascending, descending and hovering is significantly higher than that of a conventional single flying robot, and the existing inspection technology lacks a precise power consumption prediction mechanism for the shape switching and multiple motion states of the flying robot, and a return power consumption evaluation method that is in line with the actual operation scene of the flying robot is not established, so that it is difficult to accurately predict the minimum power consumption required for return during the flying process, and the remaining power cannot match the actual power consumption demand of the return, which easily causes safety risks such as equipment disconnection and damage due to falling, and seriously restricts the reliable application of the flying robot in various inspection scenes. Therefore, a flying robot inspection system and method based on Internet of Things are urgently needed. SUMMARY

[0004] The application aims to provide a flying robot inspection system and method based on Internet of Things to solve the problems in the prior art.

[0005] To achieve the above-mentioned purpose, the application provides the following technical scheme: a flying robot inspection method based on Internet of Things, the flying robot inspection method comprising the following steps: Step S1, obtaining a historical flight record of a flying robot according to a unique identifier of the flying robot, analyzing and extracting a plurality of historical flight state feature parameters, and constructing a flight state feature set according to data collection time stamps of the historical flight state feature parameters; Step S1-1, accessing a backend service database through the unique identifier of the flying robot, screening all flying robots of the same type and specification in the backend service database according to the product model and hardware configuration parameters of the flying robot, and extracting complete flight data of each flying robot, wherein the complete flight data comprises state records of each operation stage and corresponding data collection time stamps; Step S1-2: Perform structured analysis on the extracted complete flight data to separate the flight state characteristic parameters of each flying robot in different inspection stages. The flight states include ascending state, descending state, hovering state, stable flight state, crawling to flight state, and flight to crawling state. The flight state characteristic parameters include the flight speed and flight attitude angle corresponding to each flight state. Steps S1-3: Extract the total power consumption data corresponding to each flight state from the complete flight data. Calculate the duration of each state based on the time difference between the start and end data collection timestamps of each flight state. Obtain the average power consumption value of each flight state by the ratio of the total power consumption data of each flight state to the corresponding state duration. Record this as the power consumption value corresponding to the flight state. Steps S1-4: Associate and uniquely bind the flight status characteristic parameters and power consumption values ​​with the corresponding data acquisition timestamps to construct a flight status characteristic set containing multiple flying robots, multiple time nodes, and associated power consumption values ​​of each flight status. By accessing the backend service database through the unique identifier of the flying robot, complete flight data from multiple robots of the same type and specifications are filtered. Through structured parsing, flight status characteristic parameters and corresponding power consumption values ​​for each inspection stage are separated. These data are then uniquely bound to the collection timestamp to construct a standardized set of flight status characteristics. Based on the complete historical data of multiple devices, covering ascent, descent, hovering, stable flight, and two-mode switching states, the correlation between characteristic parameters such as flight speed and flight attitude angle and power consumption values ​​for each state is clarified. This solves the problems of fragmented inspection data, lack of unified standards, and lack of association with multi-state power consumption information in existing technologies, providing core data support with a unified structure and complete dimensions for power consumption prediction and return-to-base assessment.

[0006] Step S2: Create a flight state feature parameter factory to traverse and read the flight state feature set, and then use the flight state feature parameter factory to train and analyze the read flight state feature parameters to predict the first inspection power consumption value under each flight state. Step S2-1: Based on the flight state type in the flight state feature set, split the set into ascending state data subset, descending state data subset, hovering state data subset, stable flight state data subset, crawling to flight state data subset, and flight to crawl state data subset. Each data subset contains the corresponding flight state feature parameters and power consumption value. Step S2-2: Create a flight status feature parameter factory, configure a dedicated business interface for each flight status type, and establish a unique mapping relationship between each business interface and the data subset of the corresponding flight status. Step S2-3: The flight status feature parameter factory sequentially traverses each data subset according to the flight status type, extracts the flight status feature parameters in each data subset, the flight status feature parameters include flight speed and flight attitude angle, and synchronously transmits the extracted flight status feature parameters and power consumption value to the corresponding flight status business interface. Step S2-4: Each business interface has a built-in preset neural network model. After receiving the incoming flight status feature parameters, the neural network model selects the flight status feature parameters as independent variables and the synchronously incoming power consumption value as dependent variables. It analyzes and trains the independent and dependent variables. After training, it outputs the predicted power consumption value under the corresponding flight status, which is recorded as the first inspection power consumption value. Based on the state types of the flight state feature set, data subsets specific to each state are formed. By creating a flight state feature parameter factory and configuring corresponding business interfaces, the feature parameters and power consumption values ​​of each subset are input into the neural network model built into the interface. The model is trained and analyzed with the feature parameters as independent variables and the power consumption value as dependent variables, and the first inspection power consumption value is output. By adopting a combination of data classification, dedicated interfaces, and model training, a power consumption prediction mechanism based on historical common data is established for the multi-motion state and form switching scenarios of the flying and climbing robot, providing stable and reliable basic power consumption benchmark data for each flight state.

[0007] Step S3: Obtain real-time flight status characteristic parameters during flight according to the real-time operation instructions of the flying robot, set an altitude return sliding window to extract the historical flight status characteristic parameters of the current flying robot, divide the factory data according to the extracted flight status characteristic parameters, input the divided flight status characteristic parameters and real-time flight status characteristic parameters into the flight status characteristic parameter factory for training analysis, and predict the second inspection power consumption value under each flight status. Step S3-1: According to the real-time operation instructions of the flying robot, collect the real-time flight status characteristic parameters during the current inspection process, configure the window size of the altitude return sliding window, and determine the maximum number of operation instructions after one inspection by analyzing the historical inspection data in the flight status characteristic set. Set the maximum value as the sliding step size. After each inspection task is completed, slide the altitude return sliding window according to the set sliding step size, and extract the recent historical data that is not empty from the historical flight status characteristic parameters of the current flying robot itself. Specifically, it is all the flight status characteristic parameters of itself until the last inspection was completed. Step S3-2: Divide the captured flight state feature parameters according to the flight state type to obtain the recent subsets of ascending state, descending state, hovering state, stable flight state, crawling to flight state, and flight to crawl state. At the same time, classify the collected real-time flight state feature parameters according to the flight state. Step S3-3: Synchronously input the recent data subsets corresponding to each flight state and the classified real-time flight state feature parameters into the corresponding flight state business interface. The neural network model built into each business interface uses the flight state feature parameters in the recent data subset as the basic independent variable, the real-time flight state feature parameters as the supplementary independent variable, and the power consumption value in the recent data subset as the dependent variable for analysis and training. After training, the predicted power consumption value in the corresponding flight state is output and recorded as the second inspection power consumption value. Based on real-time operation commands, the current flight status characteristic parameters are collected. By configuring the size and sliding step of the altitude return sliding window, the recent historical data of the robot up to the last inspection is extracted. After being divided into status types, the data is synchronously transmitted to the corresponding business interface along with the real-time parameters. The second inspection power consumption value is output through collaborative training of the neural network model. Based on the recent historical data of the current robot, supplementary training is carried out in combination with real-time operation parameters. The sliding window is used to ensure the timeliness and non-emptiness of the data, forming a dynamic power consumption correction benchmark that fits the current flight status, improving the scenario adaptability and accuracy of power consumption prediction.

[0008] Step S4: Based on the analysis of the flight state feature set, extract the flight state feature parameters included in the return flight at different flight altitudes, construct the return flight altitude mapping table, and obtain the real-time flight altitude data of the flying robot based on the analysis of the real-time flight state feature parameters. The flight state feature parameters included in the return flight of the flying robot are obtained in real time through the return flight altitude mapping table. Step S4-1: Traverse all historical flight records in the flight state feature set, extract the flight altitude data of the return phase in each historical flight record, and synchronously correlate and extract all flight states and corresponding flight state feature parameters included in the return phase. The return phase refers to the complete process of the flying robot returning from the inspection operation altitude to the take-off and landing point. Step S4-2: Group the extracted return phase information according to the flight altitude data, count the occurrence frequency of each flight state in each group, determine the standard flight state sequence of the return phase at that flight altitude based on the occurrence frequency, use the flight altitude data as the index, use the corresponding standard flight state sequence and associated flight state feature parameters as the mapping value, establish a unique association between flight altitude and return standard flight state sequence and flight state feature parameters, and construct a return altitude mapping table; Step S4-3: Parse the real-time flight altitude data of the flying robot from the real-time flight status feature parameters, use the real-time flight altitude data as a query condition to retrieve the return altitude mapping table, and match the corresponding return standard flight status sequence and associated flight status feature parameters, which are the flight status feature parameters required for the flying robot to return at the current altitude. By traversing historical flight records in the flight state feature set, flight altitude data, flight state, and corresponding feature parameters for each return phase are extracted. The frequency of occurrence of each state is statistically analyzed by altitude to determine the standard flight state sequence. A return altitude mapping table is constructed with altitude as the index and state sequence and feature parameters as mapping values. The state feature parameters required for the current return are obtained by retrieving the mapping table through real-time flight altitude data. By statistically analyzing the state occurrence patterns of historical return data, a unique correlation between flight altitude and return state sequence is established, clarifying the core states and parameters of return at different altitudes, and providing a clear state basis for accurately calculating return power consumption.

[0009] Step S5: Analyze the flight state characteristic parameters included in the return flight, and combine them with the first inspection power consumption value and the second inspection power consumption value under each flight state to obtain the minimum inspection power consumption value required for the flight robot to return from the current flight altitude, and evaluate the inspection flight based on the remaining power of the current flight robot. Step S5-1: Extract the historical actual power consumption value corresponding to each flight state from the flight state feature set. Calculate the absolute error ratio between the first inspection power consumption value and the historical actual power consumption value for each flight state, and the absolute error ratio between the second inspection power consumption value and the historical actual power consumption value. Using the reciprocal of the two error ratios as a benchmark, calculate the proportion of the two error reciprocals to the sum of the two types of reciprocals. This proportion is the weight coefficient of the corresponding inspection power consumption value. The sum of the weight coefficients of the two inspection power consumption values ​​is 1. The formula for calculating the percentage of absolute error in the first inspection's power consumption is as follows: ; In the formula, E1 represents the percentage of the absolute error between the first inspection power consumption value and the historical actual power consumption value; P1 represents the first inspection power consumption value under each flight condition, which is predicted based on training with historical common data; P act This is represented as the historical actual power consumption value corresponding to each flight state, extracted from the flight state feature set; The formula for calculating the percentage of absolute error in the power consumption value during the second inspection is as follows: ; In the formula, E2 represents the percentage of the absolute error between the second inspection power consumption value and the historical actual power consumption value; P2 represents the second inspection power consumption value under each flight state, which is predicted based on the collaborative training of recent historical data and real-time data. The formula for calculating the weighting coefficient of the first inspection power consumption value is as follows: ; In the formula, w1 represents the weighting coefficient of the first inspection power consumption value; The formula for calculating the weighting coefficient of the power consumption value during the second inspection is as follows: ; In the formula, w2 represents the weighting coefficient of the second inspection power consumption value; and satisfies w1+w2=1; Step S5-2: For each flight state, multiply the first inspection power consumption value by its corresponding weight coefficient, and add the result of multiplying the second inspection power consumption value by its corresponding weight coefficient to obtain the comprehensive power consumption value for each flight state; according to the return standard flight state sequence, match the comprehensive power consumption value corresponding to each flight state one by one in the sequence order, and accumulate all the matched comprehensive power consumption values ​​to obtain the minimum inspection power consumption value required for the flight robot to return from the current flight altitude. The formula for calculating the total power consumption is as follows: P all =P1×E1+P2×E2; In the formula, P all This is represented as the comprehensive power consumption value corresponding to the flight state. The comprehensive power consumption value corresponding to each flight state is obtained by calculating and processing each flight state according to the absolute error ratio calculation formula, the weight coefficient calculation formula, and the comprehensive power consumption value calculation formula.

[0010] Step S5-3: Obtain the current remaining power data of the flying robot, compare the remaining power data with the minimum inspection power consumption value required for return, and generate an inspection return request and send it to the flying robot control terminal when the remaining power data does not exceed the minimum inspection power consumption value; when the remaining power data exceeds the minimum inspection power consumption value, the flying robot continues to perform the preset inspection task. Extract the historical actual power consumption value corresponding to each flight state from the flight state feature set. Calculate the absolute error ratio between the first inspection power consumption value and the historical actual power consumption value, and the absolute error ratio between the second inspection power consumption value and the historical actual power consumption value. Using the reciprocal of the two error ratios as a benchmark, calculate the proportion of the two error reciprocals to the sum of the two types of reciprocals. This proportion is the weight coefficient of the corresponding inspection power consumption value. The sum of the weight coefficients of the two inspection power consumption values ​​is 1. For each flight state, multiply the first inspection power consumption value by its corresponding weight coefficient and add the result of multiplying the second inspection power consumption value by its corresponding weight coefficient to obtain the comprehensive power consumption value for each flight state. According to the return-to-home standard flight state sequence, the comprehensive power consumption value corresponding to each flight state is matched one by one in sequence. All the matched comprehensive power consumption values ​​are accumulated to obtain the minimum inspection power consumption value required for the flight robot to return from the current flight altitude. The current remaining power data of the flight robot is obtained and compared with the minimum inspection power consumption value required for return. When the remaining power data does not exceed the minimum inspection power consumption value, an inspection return-to-home request is generated and sent to the flight robot control terminal. When the remaining power data exceeds the minimum inspection power consumption value, the flight robot continues to execute the preset inspection task. By adopting the method of allocating weights based on the inverse of the error ratio, the two types of power consumption values ​​are scientifically integrated, accurately quantifying the minimum power consumption required for returning to base at the current altitude. This effectively avoids safety risks such as equipment disconnection and damage due to insufficient power, and ensures the reliability of the flying and climbing robot's inspection operations.

[0011] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention filters complete flight data from multiple devices of the same type and specifications using a unique identifier for the flying robot. After structured parsing, it separates the characteristic parameters of each inspection state with their corresponding power consumption values, and then associates and binds them with timestamps to construct a standardized set of flight state features, covering multiple motion states and form switching scenarios. It clarifies the relationship between characteristic parameters and power consumption values, solves the problem of scattered inspection data and lack of unified standards, and provides core data support with a unified structure and complete dimensions for power consumption prediction and return assessment, ensuring the reliability and comprehensiveness of the data foundation.

[0012] 2. This invention constructs a dual power consumption prediction mechanism. It splits a data subset by a flight state feature parameter factory and configures a dedicated business interface. It uses a neural network model to train and output the first inspection power consumption value based on historical common data. Then, it combines the parameters collected by real-time operation commands and uses the altitude return sliding window to extract its own recent historical data for collaborative training to obtain the second inspection power consumption value. This achieves the combination of static benchmark and dynamic correction, which significantly improves the accuracy and scenario adaptability of power consumption prediction in multi-state and form switching scenarios.

[0013] 3. This invention constructs a return altitude mapping table by statistically analyzing historical return data, clearly defining the standard state sequence corresponding to different flight altitudes. Then, it integrates the power consumption values ​​of dual inspections by allocating weights based on the inverse of the error ratio, accurately calculating the minimum power consumption value required for return. By comparing the remaining power, it generates a return request or allows continued inspection, effectively avoiding the risk of loss of connection and crash damage to the climbing robot due to insufficient power, significantly improving the safety and reliability of inspection operations, and ensuring the stable application of the equipment in complex scenarios. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an IoT-based inspection method for flying robots according to the present invention. Figure 2 This is a schematic diagram of the structure of an IoT-based flying robot inspection system according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example 1: As Figure 1 As shown, the present invention provides a technical solution, a flight robot inspection method based on the Internet of Things, the flight robot inspection method including the following steps: Step S1: Obtain the historical flight records of the flying robot based on its unique identifier, analyze and extract multiple historical flight state feature parameters, and construct a flight state feature set based on the data acquisition timestamps of each historical flight state feature parameter. Step S1-1: Access the backend service database through the unique identifier of the flying robot. Based on the product model and hardware configuration parameters of the flying robot, filter all flying robots of the same type and specifications in the backend service database and extract the complete flight data of each flying robot. The complete flight data includes the status records of each operating stage and the corresponding data collection timestamp. Step S1-2: Perform structured analysis on the extracted complete flight data to separate the flight state characteristic parameters of each flying robot in different inspection stages. The flight states include ascending state, descending state, hovering state, stable flight state, crawling to flight state, and flight to crawling state. The flight state characteristic parameters include the flight speed and flight attitude angle corresponding to each flight state. Steps S1-3: Extract the total power consumption data corresponding to each flight state from the complete flight data. Calculate the duration of each state based on the time difference between the start and end data collection timestamps of each flight state. Obtain the average power consumption value of each flight state by the ratio of the total power consumption data of each flight state to the corresponding state duration. Record this as the power consumption value corresponding to the flight state. Steps S1-4: Associate and uniquely bind the flight status characteristic parameters and power consumption values ​​with the corresponding data acquisition timestamps to construct a flight status characteristic set containing multiple flying robots, multiple time nodes, and associated power consumption values ​​of each flight status. In practice, the unique identifier of the climbing robot is used to accurately connect to the backend service database. When filtering complete flight data of devices of the same specifications, it is necessary to ensure that the product model and hardware configuration parameters are completely consistent. During the structured parsing process, the feature parameters of various flight states must be completely separated. The timestamp association binding operation must ensure the uniqueness of each data node to avoid data confusion between different robots or at different times, which would affect the standardization of the feature set.

[0017] Step S2: Create a flight state feature parameter factory to traverse and read the flight state feature set, and then use the flight state feature parameter factory to train and analyze the read flight state feature parameters to predict the first inspection power consumption value under each flight state. Step S2-1: Based on the flight state type in the flight state feature set, split the set into ascending state data subset, descending state data subset, hovering state data subset, stable flight state data subset, crawling to flight state data subset, and flight to crawl state data subset. Each data subset contains the corresponding flight state feature parameters and power consumption value. Step S2-2: Create a flight status feature parameter factory, configure a dedicated business interface for each flight status type, and establish a unique mapping relationship between each business interface and the data subset of the corresponding flight status. Step S2-3: The flight status feature parameter factory sequentially traverses each data subset according to the flight status type, extracts the flight status feature parameters in each data subset, the flight status feature parameters include flight speed and flight attitude angle, and synchronously transmits the extracted flight status feature parameters and power consumption value to the corresponding flight status business interface. Step S2-4: Each business interface has a built-in preset neural network model. After receiving the incoming flight status feature parameters, the neural network model selects the flight status feature parameters as independent variables and the synchronously incoming power consumption value as dependent variables. It analyzes and trains the independent and dependent variables. After training, it outputs the predicted power consumption value under the corresponding flight status, which is recorded as the first inspection power consumption value. In practical implementation, the creation of the flight status feature parameter factory needs to be compatible with all preset flight status types. The mapping relationship between the dedicated business interface and the data subset needs to be fixed in advance. When traversing the data subset, it should be executed sequentially according to the status type to avoid cross-reading. After receiving the parameters, the neural network model should be trained strictly with the flight status feature parameters as independent variables and the power consumption value as dependent variables to ensure that the output first inspection power consumption value is accurately matched with the corresponding flight status.

[0018] Step S3: Obtain real-time flight status characteristic parameters during flight according to the real-time operation instructions of the flying robot, set an altitude return sliding window to extract the historical flight status characteristic parameters of the current flying robot, divide the factory data according to the extracted flight status characteristic parameters, input the divided flight status characteristic parameters and real-time flight status characteristic parameters into the flight status characteristic parameter factory for training analysis, and predict the second inspection power consumption value under each flight status. Step S3-1: According to the real-time operation instructions of the flying robot, collect the real-time flight status characteristic parameters during the current inspection process, configure the window size of the altitude return sliding window, and determine the maximum number of operation instructions after one inspection by analyzing the historical inspection data in the flight status characteristic set. Set the maximum value as the sliding step size. After each inspection task is completed, slide the altitude return sliding window according to the set sliding step size, and extract the recent historical data that is not empty from the historical flight status characteristic parameters of the current flying robot itself. Specifically, it is all the flight status characteristic parameters of itself until the last inspection was completed. Step S3-2: Divide the captured flight state feature parameters according to the flight state type to obtain the recent subsets of ascending state, descending state, hovering state, stable flight state, crawling to flight state, and flight to crawl state. At the same time, classify the collected real-time flight state feature parameters according to the flight state. Step S3-3: Synchronously input the recent data subsets corresponding to each flight state and the classified real-time flight state feature parameters into the corresponding flight state business interface. The neural network model built into each business interface uses the flight state feature parameters in the recent data subset as the basic independent variable, the real-time flight state feature parameters as the supplementary independent variable, and the power consumption value in the recent data subset as the dependent variable for analysis and training. After training, the predicted power consumption value in the corresponding flight state is output and recorded as the second inspection power consumption value. In practice, the size of the altitude return sliding window needs to be reasonably configured in combination with the overall scale of historical inspection data. The determination of the sliding step size should be based solely on the maximum number of operation commands after one inspection is completed. When extracting recent historical data, it should be limited to the current flight robot's own historical records. When real-time parameters and recent data subsets are passed to the business interface, they should be classified and corresponded according to flight status. During collaborative training, the supplementary and corrective role of real-time parameters on historical data should be emphasized.

[0019] Step S4: Based on the analysis of the flight state feature set, extract the flight state feature parameters included in the return flight at different flight altitudes, construct the return flight altitude mapping table, and obtain the real-time flight altitude data of the flying robot based on the analysis of the real-time flight state feature parameters. The flight state feature parameters included in the return flight of the flying robot are obtained in real time through the return flight altitude mapping table. Step S4-1: Traverse all historical flight records in the flight state feature set, extract the flight altitude data of the return phase in each historical flight record, and synchronously correlate and extract all flight states and corresponding flight state feature parameters included in the return phase. The return phase refers to the complete process of the flying robot returning from the inspection operation altitude to the take-off and landing point. Step S4-2: Group the extracted return phase information according to the flight altitude data, count the occurrence frequency of each flight state in each group, determine the standard flight state sequence of the return phase at that flight altitude based on the occurrence frequency, use the flight altitude data as the index, use the corresponding standard flight state sequence and associated flight state feature parameters as the mapping value, establish a unique association between flight altitude and return standard flight state sequence and flight state feature parameters, and construct a return altitude mapping table; Step S4-3: Parse the real-time flight altitude data of the flying robot from the real-time flight status feature parameters, use the real-time flight altitude data as a query condition to retrieve the return altitude mapping table, and match the corresponding return standard flight status sequence and associated flight status feature parameters, which are the flight status feature parameters required for the flying robot to return at the current altitude. In practice, the historical flight records must be thoroughly covered to cover all return phase data. When extracting flight altitude and state feature parameters for the return phase, a one-to-one correspondence must be maintained. After grouping by altitude, the number of occurrences of each state must be counted to ensure the integrity of the statistical range. The index items and mapping values ​​of the return altitude mapping table must be uniquely associated. When searching by real-time flight altitude, the corresponding standard flight state sequence must be quickly matched to ensure the efficiency of obtaining the state feature parameters required for the return phase.

[0020] Step S5: Analyze the flight state characteristic parameters included in the return flight, and combine them with the first inspection power consumption value and the second inspection power consumption value under each flight state to obtain the minimum inspection power consumption value required for the flight robot to return from the current flight altitude, and evaluate the inspection flight based on the remaining power of the current flight robot. Step S5-1: Extract the historical actual power consumption value corresponding to each flight state from the flight state feature set. Calculate the absolute error ratio between the first inspection power consumption value and the historical actual power consumption value for each flight state, and the absolute error ratio between the second inspection power consumption value and the historical actual power consumption value. Using the reciprocal of the two error ratios as a benchmark, calculate the proportion of the two error reciprocals to the sum of the two types of reciprocals. This proportion is the weight coefficient of the corresponding inspection power consumption value. The sum of the weight coefficients of the two inspection power consumption values ​​is 1. Step S5-2: For each flight state, multiply the first inspection power consumption value by its corresponding weight coefficient, and add the result of multiplying the second inspection power consumption value by its corresponding weight coefficient to obtain the comprehensive power consumption value for each flight state; according to the return standard flight state sequence, match the comprehensive power consumption value corresponding to each flight state one by one in the sequence order, and accumulate all the matched comprehensive power consumption values ​​to obtain the minimum inspection power consumption value required for the flight robot to return from the current flight altitude. Step S5-3: Obtain the current remaining power data of the flying robot, compare the remaining power data with the minimum inspection power consumption value required for return, and generate an inspection return request and send it to the flying robot control terminal when the remaining power data does not exceed the minimum inspection power consumption value; when the remaining power data exceeds the minimum inspection power consumption value, the flying robot continues to perform the preset inspection task. In practice, the extracted historical actual power consumption values ​​must correspond precisely to each flight state. When calculating the error ratio, the historical actual power consumption value must be used as the benchmark. The calculation of the weighting coefficient must strictly follow the proportional allocation rule of the inverse of the error ratio. The calculation of the comprehensive power consumption value must be performed separately for each flight state. The accumulation process must strictly follow the order of the return standard flight state sequence. The comparison between the remaining power and the minimum inspection power consumption value must be performed in real time to ensure that the results of the return request generation or the determination of the inspection task continuation are accurate and reliable.

[0021] Example 2, as Figure 2 As shown, the present invention provides an Internet of Things-based flying robot inspection system, which includes a historical data processing module, a first power consumption prediction module, a second power consumption prediction module, a return status matching module, and an inspection power consumption assessment module. The historical data processing module is used to obtain complete flight data of flying robots of the same type and specifications through the unique identifier of the flying robot, and construct a flight state feature set; the first power consumption prediction module is used to divide and train the flight state feature set to predict the first inspection power consumption value under each flight state; the second power consumption prediction module is used to extract data through an altitude return sliding window, and combine it with real-time flight state feature parameters for training to predict the second inspection power consumption value under each flight state; the return state matching module is used to construct a return altitude mapping table based on the flight state feature set, and match the return flight state feature parameters corresponding to the current flight altitude; the inspection power consumption assessment module is used to calculate the minimum inspection power consumption value required for return, and combine it with the remaining power to assess whether the inspection flight should continue. The historical data processing module includes a data filtering and extraction unit and a feature set construction unit. The data filtering and extraction unit is used to access the backend service database to filter and extract complete flight data of each flying robot and its corresponding flight status feature parameters and power consumption values. The feature set construction unit is used to uniquely bind the flight status feature parameters and power consumption values ​​with the corresponding data collection timestamps to form a flight status feature set. The first power consumption prediction module includes a data subset partitioning unit and a model training prediction unit; the data subset partitioning unit is used to split the flight state feature set into corresponding data subsets according to the flight state type; the model training prediction unit is used to train the data subsets through the neural network model built into the business interface and output the first inspection power consumption value. The second power consumption prediction module includes a sliding window interception unit and a real-time data training unit; the sliding window interception unit is used to configure the window size and sliding step of the altitude return sliding window, and intercept the recent historical flight status feature parameters of the current flight robot; the real-time data training unit is used to input the recent data subset and the real-time flight status feature parameters into the business interface for training, and output the second inspection power consumption value; The return-to-home status matching module includes a return-to-home mapping construction unit and an altitude status matching unit. The return-to-home mapping construction unit is used to extract flight altitude data, flight status and feature parameters from historical return-to-home phases to construct a return-to-home altitude mapping table. The altitude status matching unit is used to retrieve the return-to-home altitude mapping table using real-time flight altitude data to obtain the flight status feature parameters required for return-to-home at the current altitude. The inspection power consumption assessment module includes a return power consumption calculation unit and a remaining power consumption assessment unit. The return power consumption calculation unit is used to determine the weighting coefficients of the first inspection power consumption value and the second inspection power consumption value, and to calculate the comprehensive power consumption value for each flight state and the minimum inspection power consumption value for return. The remaining power consumption assessment unit is used to compare the remaining power consumption data with the minimum inspection power consumption value for return, and to generate an inspection return request or allow the continuation of the inspection task.

[0022] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

Claims

1. An inspection method for flying robots based on the Internet of Things, characterized in that: The flying robot inspection method comprises the following steps: Step S1, obtaining the historical flight record of the flying robot according to the unique identifier of the flying robot, analyzing and extracting a plurality of historical flight state characteristic parameters, and constructing a flight state characteristic set according to the data collection time stamps of the historical flight state characteristic parameters; Step S2, creating a flight state characteristic parameter factory to read the flight state characteristic set, and after training and analyzing the read flight state characteristic parameters by the flight state characteristic parameter factory, predicting a first inspection power consumption value in each flight state; Step S3, obtaining real-time flight state characteristic parameters in the flight process according to the real-time operation instruction of the flying robot, setting a height return sliding window to intercept the historical flight state characteristic parameters of the current flying robot, dividing the intercepted flight state characteristic parameters into factory data according to the flight state characteristic parameters, inputting the divided flight state characteristic parameters and the real-time flight state characteristic parameters into the flight state characteristic parameter factory for training and analysis, and predicting a second inspection power consumption value in each flight state; Step S4, analyzing and extracting flight state characteristic parameters included in different flight heights based on the flight state characteristic set, constructing a return height mapping table, analyzing the real-time flight height data of the flying robot according to the real-time flight state characteristic parameters, and obtaining the flight state characteristic parameters included in the return of the flying robot in real time through the return height mapping table; Step S5, combining the flight state characteristic parameters included in the return with the first inspection power consumption value and the second inspection power consumption value in each flight state to analyze and obtain the minimum inspection power consumption value required for the flying robot to return from the current flight height, and evaluating the inspection flight according to the remaining power of the current flying robot. 2.The flight robot patrol method based on the Internet of Things according to claim 1, characterized in that: The specific steps of step S1 are as follows: Step S1-1, accessing the backend service database through the unique identifier of the flying robot, filtering all flying robots of the same type and specification in the backend service database according to the product model and hardware configuration parameters of the flying robot, and extracting complete flight data of each flying robot, wherein the complete flight data includes state records and corresponding data collection time stamps in each operating phase; Step S1-2, structurally analyzing the extracted complete flight data to separate the flight state characteristic parameters of each flying robot in different inspection phases, wherein the flight states include ascending state, descending state, hovering state, steady flight state, climbing to flight state and flight to climbing state, and the flight state characteristic parameters include flight speed and flight attitude angle corresponding to each flight state; Step S1-3, extracting the total power consumption data corresponding to each flight state from the complete flight data, calculating the time difference to obtain the state duration based on the start data collection time stamp and the end data collection time stamp of each flight state, and obtaining the average power consumption value of each flight state by the ratio of the total power consumption data of each flight state to the corresponding state duration, which is recorded as the power consumption value corresponding to the flight state; Step S1-4, the flight state characteristic parameters and the power consumption value are uniquely bound with the corresponding data collection time stamp, and a flight state characteristic set containing multiple flight robots, multiple time nodes, and associated flight state power consumption values is constructed. 3.The flight robot patrol method based on the Internet of Things according to claim 2, characterized in that: The specific steps of step S2 are as follows: Step S2-1, based on the flight state type in the flight state characteristic set, the set is split into an ascending state data subset, a descending state data subset, a hovering state data subset, a stable flight state data subset, a crawling-to-flight state data subset, a flight-to-crawling state data subset, and each data subset contains flight state characteristic parameters and power consumption values corresponding to the flight state; Step S2-2, a flight state characteristic parameter factory is created, and dedicated business interfaces are configured for each flight state type, and each business interface has a unique mapping relationship with the data subset corresponding to the flight state.

4. The method of claim 3, wherein the method further comprises: In step S2, it also includes: Step S2-3, the flight state characteristic parameter factory traverses each data subset in turn according to the flight state type, extracts the flight state characteristic parameters in each data subset, and the flight state characteristic parameters include flight speed and flight attitude angle, and synchronously transmits the extracted flight state characteristic parameters and power consumption values into the business interface corresponding to the flight state; Step S2-4, each business interface has a preset neural network model, which selects flight state characteristic parameters as independent variables and synchronously transmitted power consumption values as dependent variables after receiving the transmitted flight state characteristic parameters, analyzes and trains the independent variables and dependent variables, and outputs the predicted power consumption value under the corresponding flight state after training, which is recorded as the first inspection power consumption value.

5. The method of claim 4, wherein the method further comprises: The specific steps of step S3 are as follows: Step S3-1, according to the real-time operation instruction of the flight robot, the real-time flight state characteristic parameters in the current inspection process are collected, the window size of the height return sliding window is configured, the maximum value of the operation instruction number after one inspection is determined by analyzing the historical inspection data in the flight state characteristic set, and the maximum value is set as the sliding step length. After each inspection task is completed, the height return sliding window is slid by the set sliding step length, and the non-empty recent historical data is intercepted from the historical flight state characteristic parameters of the current flight robot itself, specifically all flight state characteristic parameters until the completion of the last inspection; Step S3-2, the intercepted flight state characteristic parameters are divided according to the flight state type to obtain an ascending state recent subset, a descending state recent subset, a hovering state recent subset, a stable flight state recent subset, a crawling-to-flight state recent subset, and a flight-to-crawling state recent subset, and the collected real-time flight state characteristic parameters are classified according to the flight state. Step S3-3, synchronously transmit the recent data subset corresponding to each flight state and the classified real-time flight state feature parameter into the business interface of the corresponding flight state, the neural network model built in each business interface takes the flight state feature parameter in the recent data subset as the basic independent variable, takes the real-time flight state feature parameter as the supplementary independent variable, and takes the power consumption value in the recent data subset as the dependent variable for analysis and training, and outputs the predicted power consumption value under the corresponding flight state after the training is completed, which is recorded as the second inspection power consumption value. 6.The method of claim 5, wherein the method further comprises: The specific steps of the step S4 are as follows: Step S4-1, traverse all the historical flight records in the flight state feature set, extract the flight height data of the homeward journey stage in each historical flight record, synchronously associate and extract all the flight states contained in the homeward journey stage and the corresponding flight state feature parameters, the homeward journey stage refers to the complete process of returning to the take-off point from the inspection operation height of the flying robot; Step S4-2, group the extracted homeward journey stage information according to the flight height data, count the number of occurrences of each flight state in each group, determine the standard flight state sequence of the homeward journey stage at the flight height according to the number of occurrences, take the flight height data as the index item, take the corresponding standard flight state sequence and the associated flight state feature parameters as the mapping value, establish the unique association of the flight height with the standard flight state sequence of the homeward journey and the flight state feature parameters, and construct the homeward journey height mapping table; Step S4-3, the real-time flight height data of the flying robot is parsed from the real-time flight state feature parameters, the real-time flight height data is taken as the query condition, the homeward journey height mapping table is searched, and the corresponding standard flight state sequence of the homeward journey and the associated flight state feature parameters are matched, which are the flight state feature parameters required for the homeward journey at the current height of the flying robot.

7. The method of claim 6, wherein the method further comprises: The specific steps of the step S5 are as follows: Step S5-1, extract the historical actual power consumption value corresponding to each flight state from the flight state feature set, respectively calculate the absolute error proportion of the first inspection power consumption value and the historical actual power consumption value under each flight state, and the absolute error proportion of the second inspection power consumption value and the historical actual power consumption value, take the reciprocal of the two error proportions as the benchmark, respectively calculate the proportion of the two error reciprocals in the total sum of the two reciprocals, and the proportion is the weight coefficient of the corresponding inspection power consumption value, and the sum of the weight coefficients of the two inspection power consumption values is 1; Step S5-2, for each flight state, multiply the first inspection power consumption value by its corresponding weight coefficient, and add the result of multiplying the second inspection power consumption value by its corresponding weight coefficient to obtain the comprehensive power consumption value under each flight state; According to the standard flight state sequence of the homeward journey, the comprehensive power consumption value corresponding to each flight state is matched in sequence, all the matched comprehensive power consumption values are accumulated, and the minimum inspection power consumption value required for the flying robot to return from the current flight height is obtained. Step S5-3, obtaining the current remaining power data of the flying robot, comparing the remaining power data with the minimum inspection power consumption value required for returning, when the remaining power data does not exceed the minimum inspection power consumption value, generating a return request for inspection and sending it to the flying robot control terminal; when the remaining power data exceeds the minimum inspection power consumption value, the flying robot continues to perform the preset inspection task.

8. An Internet of Things-based flying robot inspection system applied to the Internet of Things-based flying robot inspection method of any one of claims 1-7. The flying robot inspection system comprises a historical data processing module, a first power consumption value prediction module, a second power consumption value prediction module, a return state matching module and an inspection power evaluation module. The historical data processing module is configured to obtain complete flight data of flying robots of the same type and specification through a unique identifier of the flying robot, and construct a flight state feature set; the first power consumption value prediction module is configured to divide and analyze the flight state feature set, and predict the first inspection power consumption value under each flight state; the second power consumption value prediction module is configured to intercept data through a height return sliding window, and combine real-time flight state feature parameters for training, to predict the second inspection power consumption value under each flight state; the return state matching module is configured to construct a return height mapping table based on the flight state feature set, and match the return flight state feature parameters corresponding to the current flight height; The inspection power evaluation module is configured to calculate the minimum inspection power consumption value required for returning, and evaluate whether the inspection flight continues in combination with the remaining power.

9. The flying robot inspection system based on the Internet of Things according to claim 8, characterized in that: The historical data processing module comprises a data screening and extraction unit and a feature set construction unit; the data screening and extraction unit is configured to access a backend service database, and screen and extract complete flight data of each flying robot and corresponding flight state feature parameters and power consumption values; the feature set construction unit is configured to uniquely bind the flight state feature parameters, power consumption values and corresponding data collection time stamps, to form a flight state feature set; The first power consumption value prediction module comprises a data subset division unit and a model training and prediction unit; the data subset division unit is configured to split the flight state feature set into corresponding data subsets according to flight state types; the model training and prediction unit is configured to train the data subsets through a neural network model built in a business interface, and output the first inspection power consumption value; The second power consumption value prediction module comprises a sliding window interception unit and a real-time data training unit; the sliding window interception unit is configured to configure the window size and sliding step of the height return sliding window, and intercept recent historical flight state feature parameters of the current flying robot itself; the real-time data training unit is configured to input the recent data subsets and real-time flight state feature parameters into the business interface for training, and output the second inspection power consumption value.

10. The flying robot inspection system based on the Internet of Things according to claim 8, characterized in that: The return state matching module comprises a return mapping construction unit and a height state matching unit; the return mapping construction unit is configured to extract flight height data, flight state and characteristic parameters of a historical return phase, and construct a return height mapping table; the height state matching unit is configured to search the return height mapping table through real-time flight height data, and obtain flight state characteristic parameters required for return at a current height; The inspection power consumption evaluation module comprises a return power consumption calculation unit and a residual power evaluation unit; the return power consumption calculation unit is configured to determine a weight coefficient of a first inspection power consumption value and a second inspection power consumption value, calculate a comprehensive power consumption value of each flight state and a minimum inspection power consumption value of return; The residual power evaluation unit is configured to compare residual power data with the minimum inspection power consumption value of return, and generate an inspection return request or allow the inspection task to continue to be executed.

Images