Robot remote control inspection method and system

Abstract

The invention discloses a robot remote control inspection method and system, and belongs to the technical field of inspection, and the method comprises the steps: responding to a manual remote control instruction, and constructing an inspection plan according to one or more target sites selected by a manual remote control end; robot information, station information, charging point information and historical inspection information are acquired, an environmental resistance coefficient is updated adaptively based on a meta-learning mechanism, and a safe electric quantity constraint target is set; in combination with a path planning mode, dynamically generating a driving path with safe electric quantity as a constraint; a robot driving mode is set, and after the robot is controlled to move, the motion state of the robot is adjusted in a self-adaptive mode based on the dynamic comparison result of the robot remaining electric quantity and the safety electric quantity constraint; the robot information, the station information, the inspection data and the alarm information are acquired and stored in real time until the inspection plan is completed, and the robot control right is transferred, so that the balance between the inspection efficiency and the flexibility is realized, and the adaptability to complex scenes is effectively improved.

Description

Technical Field

[0001] This invention relates to a robot remote-controlled inspection method and system, belonging to the field of inspection technology. Background Technology

[0002] As a core piece of equipment to replace manual labor in high-risk, high-frequency inspection tasks, inspection robots can monitor the status of stations in real time, capture and handle potential faults in a timely manner, and play an irreplaceable positive role in avoiding safety accidents and ensuring the stable operation of stations. Therefore, they have gained widespread attention and application in the process of industrial automation upgrading.

[0003] Chinese Patent CN117782084A discloses a path planning system and method for a converter station inspection robot, including a mechanical structure module, a power system module providing motion for the robot, a control system module for computation, path planning, and remote control, and a communication module for transmitting communication data. The mechanical structure module includes a chassis serving as a motion carrier, sensors for navigation, detecting objects within the station, and distance measurement, and a camera for capturing and recording abnormal equipment during inspection. The inspection robot is mounted on the chassis, and the sensors and camera are mounted on the robot. The power system module includes a battery for power supply and a motor for providing power. The control system module includes a computation control unit, a path planning algorithm unit, and a remote control unit. The communication module includes communication equipment and a data transmission unit.

[0004] Although existing technologies determine the optimal path for inspection robots through path planning, the decision-making dimension is singular and lacks adaptability to environments with uncertainty. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a robot remote control inspection method and system that transfers robot control to a human remote control terminal, supports flexible switching between multiple driving modes and path planning modes, comprehensively considers robot performance, charging needs, and the inspection requirements of different stations to plan paths and convert them into motion commands, thereby ensuring the completion of the inspection plan.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for remote-controlled inspection of robots includes:

[0008] S1. Responding to manual remote control commands, construct an inspection plan based on one or more target stations selected by the manual remote control terminal;

[0009] S2. Obtain robot information, site information, charging point information, and historical inspection information, and adaptively update the environmental resistance coefficient based on the meta-learning mechanism, and set a safe power constraint target;

[0010] S3. Combined with the path planning mode, dynamically generate driving paths constrained by safe battery level;

[0011] S4. Set the robot driving mode, and after controlling the robot to move, adaptively adjust the robot's movement state based on the dynamic comparison result between the robot's remaining power and the safe power constraint.

[0012] S5. Acquire and store robot information, site information, inspection data and alarm information in real time until the inspection plan is completed and control of the robot is handed over.

[0013] Furthermore, the adaptive update of the environmental resistance coefficient based on the meta-learning mechanism includes:

[0014] The driving data corresponding to the connection path between each target station and the next numbered target station in each inspection task is obtained. After data cleaning, a historical inspection information database is constructed. The driving data includes discretized labels of the average slope and ground material type corresponding to each connection path.

[0015] Based on the physical model, the actual environmental resistance coefficient is calculated by inversely calculating the ratio of the actual driving power consumption to the product of the connection path length and the robot's energy consumption per unit distance.

[0016] A lightweight neural network model is constructed, with discretized labels of average slope and ground material type as input and environmental resistance coefficient as output. It is initially trained based on a historical inspection information database to minimize the difference between the predicted value and the actual environmental resistance coefficient, thus obtaining an initial environmental resistance coefficient model. Newly generated valid data samples are automatically added to the training queue for incremental updates. A meta-learning mechanism is introduced to quickly generate an initial prediction model for new terrain to overcome the cold start problem.

[0017] Based on the latest updated environmental resistance coefficient model, the corresponding environmental resistance coefficient is determined by discretizing the labels according to the average slope and ground material type of each connection path.

[0018] Furthermore, setting safe power constraint targets includes:

[0019] Using the robot's current position and each target station as nodes, construct a fully connected graph and generate a set of worst-case paths and direct-connection path distances;

[0020] Obtain the estimated driving distance between each target station and the next numbered target station from the station information;

[0021] The environmental drag coefficients and distances of the direct paths between each target station and the next target station in the worst-case path are obtained. The estimated driving energy consumption is obtained by calculating the product and summing them.

[0022] The average energy consumption of the robot at a single station is obtained. The estimated energy consumption at a single station is obtained by multiplying the number of selected target stations by the energy consumption of the robot at a single station.

[0023] The estimated total energy consumption is obtained by summing the estimated driving energy consumption and the estimated stationary energy consumption.

[0024] A preset safety buffer power level is set, with the total estimated energy consumption plus the safety buffer power level being the safe power level.

[0025] Furthermore, the set of worst-case path and direct path distances is generated as follows:

[0026] S310: Obtain the estimated driving distance between each target station from the station information, and use it as the weight of the edge between each pair of nodes in the fully connected graph; obtain the shortest path between each pair of nodes through the path planning algorithm;

[0027] S320: Using a greedy strategy, starting from the robot's current position S0, select the target station Sf1 that is furthest from the robot's current position, and move the robot from its current position S0 to the target station Sf1. f1 The distance D[S0][S f1 Add the direct path distance to the set and mark the target site S. f1 Sites already visited;

[0028] S330: Among the target sites that have never been visited, find the one furthest from the target site S. f1 The farthest station S f2 Target site S f1 To the target site S f2 Distance D[S f1 ][S f2 Add the direct path distance to the set and mark the target site S. f2 Sites already visited;

[0029] S340: Repeat step S330 until all target sites are visited, and obtain the worst-case path and the set of direct path distances;

[0030] Furthermore, combining the path planning model, the generated driving path, constrained by safe battery power, includes:

[0031] Select one of the preset path planning modes as the robot's path planning mode;

[0032] When set to uplink / downlink mode, the station number or physical coordinates are obtained based on the station information. The uplink or downlink sequence of the target station in the inspection plan is preset. When the path planning constraints are met, the robot's travel path is determined.

[0033] When set to optimal path mode, the robot's travel path is determined by calculating multi-dimensional costs, constructing a minimum cost path planning model based on the A* algorithm to obtain the optimal path, and integrating safety power constraints.

[0034] When set to free mode, the inspection path is manually set and a feasibility check is performed at each step. When the check passes and the path planning constraints are met, the robot's travel path is determined.

[0035] Furthermore, calculating multi-dimensional costs includes:

[0036] Calculate site priority by calculating the difference between the preset highest priority and the site priority to obtain the site priority cost;

[0037] Based on the robot information, the total battery power and remaining battery power of the robot are extracted and the difference is calculated to obtain the degree of battery consumption; based on the station information and charging point information, the charging travel distance between each station and the charging point is obtained, and the charging convenience is obtained by calculating the ratio of the charging travel distance to the farthest distance that the robot can travel when fully charged;

[0038] The cost of charging demand is obtained by multiplying the power consumption level and the charging convenience index.

[0039] Retrieve all records of the distance between the current site and the next target site, and take the average value as the distance cost.

[0040] Furthermore, calculating site priorities includes:

[0041] Based on the historical inspection information, obtain the frequency of fault occurrence within a preset period of the site and the frequency of replanning of direct connection paths within a preset period of the path.

[0042] The site priority is obtained by weighted summing of the failure frequency within a preset period and the direct path replanning frequency within the preset period.

[0043] Furthermore, a minimum cost path planning model based on the A* algorithm is constructed to obtain the optimal path, and safety power constraints are integrated, including:

[0044] S410: Normalize and sum the site priority cost, charging demand cost, and distance cost of the sites that have been inspected, and calculate the actual cost already traveled.

[0045] S420: Perform a safety power constraint verification. If the verification fails, backtrack to step S410; if the verification passes, proceed to step S430.

[0046] S430: For uninspected stations, calculate their own priority and charging cost, as well as the distance cost to other uninspected stations, and sum them up to obtain the remaining cost.

[0047] S440: Calculate the total estimated cost by summing the actual cost already incurred and the remaining cost; perform a safety power constraint check, and prioritize the site that meets the safety power constraint and always minimizes the estimated cost.

[0048] S450: Repeat steps S410-S440 until all target stations are traversed, and finally determine the driving path with the lowest total cost from all feasible paths that meet the safe power constraints.

[0049] Furthermore, adaptive adjustment of the robot's motion state includes:

[0050] The robot determines whether its remaining battery power exceeds the preset start threshold. If it does, the robot is controlled to travel along a predetermined path. Otherwise, it goes to the charging station or issues a low battery warning.

[0051] During the robot inspection process, the estimated total energy consumption of the current inspection plan is dynamically calculated, and the operable power threshold is adjusted simultaneously; based on the comparison results of the robot's remaining power and operable power threshold obtained in real time, the robot's motion state is adjusted.

[0052] A robot remote-controlled inspection system includes: a data acquisition module, a data storage module, a planning and control module, a communication module, and a display module;

[0053] The data acquisition module includes:

[0054] The positioning unit is used to acquire the robot's physical coordinates in real time;

[0055] A multimodal sensor integration unit is used to acquire robot information, ground slope, and inspection data;

[0056] The data storage module is used to store the information acquired by the data acquisition module, including site information, charging point information, historical inspection information, and intelligent learning model parameters.

[0057] The planning control module includes:

[0058] The intelligent computing unit is used to calculate the estimated power consumption, site priority, multi-dimensional costs, and plan the driving route, and dynamically update the environmental resistance coefficient prediction model based on historical inspection data.

[0059] The data conversion unit is used to convert the data from the computing unit into executable motion commands and to convert manual control signals into drive signals.

[0060] The configuration unit is used to switch control permissions between the management system and the manual remote control terminal, and to set the robot's driving mode.

[0061] The communication module is used for real-time transmission of data and instructions;

[0062] The display module is used to display information and receive manual operations, providing functions such as setting inspection plans, switching permissions and driving modes, planning routes, controlling robots, and displaying alarm information.

[0063] The beneficial effects of this invention are as follows: by comprehensively considering robot information, site information, charging point information, and historical inspection information, and combining multi-dimensional cost calculation to ensure range safety, it provides multiple driving modes and path planning modes to achieve a balance between efficiency and flexibility. Real-time data updates support continuous optimization, effectively improving inspection accuracy, emergency response efficiency, and adaptability to complex scenarios, and ensuring the safe and efficient execution of inspection tasks. Attached Figure Description

[0064] Figure 1 Flowchart of the robot remote-controlled inspection method;

[0065] Figure 2 This is a flowchart illustrating the calculation of the required cost for this invention.

[0066] Figure 3 This is a flowchart of the minimum cost path planning based on the A* algorithm of this invention;

[0067] Figure 4 This is a structural diagram of the robot remote-controlled inspection system of the present invention. Detailed Implementation

[0068] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0069] refer to Figures 1 to 3 The diagram illustrates a specific implementation of a robot remote-controlled inspection method according to the present invention, comprising the following steps:

[0070] S1: In response to manual remote control commands, the robot stops its current task and saves the current task status, generates a list of completed inspection sites and a list of uninspected sites, and obtains the coordinates of the interruption location through GPS or SLAM technology; it constructs an inspection plan based on one or more target sites selected by the manual remote control terminal; each record in the site list includes a unique site identifier and an inspection timestamp.

[0071] S2: Acquire robot information, site information, charging point information, and historical inspection information, and adaptively update the environmental resistance coefficient based on the meta-learning mechanism, and set a safe power constraint target;

[0072] S3: Combines path planning mode to dynamically generate driving paths constrained by safe battery level;

[0073] S4: Set the robot's driving mode to manual remote control driving mode or automatic driving mode; when set to manual remote control driving mode, the robot's movement direction and speed are controlled by the manual control terminal; when set to automatic driving mode, the robot drives autonomously according to the driving path instructions without the need for real-time manual operation; after sending the motion command to control the robot to start moving, the robot's motion state is adaptively adjusted based on the dynamic comparison result between the robot's remaining power and the safe power constraint.

[0074] S5: Real-time acquisition and storage of robot information, site information, inspection data, and alarm information, updated to the database for priority calculation and model optimization; after the inspection plan is completed, robot control is transferred, inspection breakpoint information is called, and the robot is controlled to resume the inspection task from the inspection breakpoint; among them, the inspection data includes at least the detection parameters of each site and the distance between each site and the next target site, and the alarm information includes at least the site abnormal alarm and the robot fault tag.

[0075] Furthermore, the adaptive update of the environmental resistance coefficient based on the meta-learning mechanism includes:

[0076] The driving data corresponding to the connection path between each target station and the next numbered target station in each inspection task is obtained from the historical inspection information. After data cleaning, a historical inspection information database is constructed. The driving data includes discretized labels of actual driving power consumption, average slope and ground material type for each connection path.

[0077] Based on the physical model, the actual environmental drag coefficient is calculated as a monitoring signal by inversely using the ratio of the actual power consumption during travel to the product of the connection path length and the robot's energy consumption per unit distance. The expression is as follows:

[0078]

[0079] in, This is the actual environmental drag coefficient. This represents the actual power consumption during driving. This is the length of the connection path. Energy consumption per unit distance for the robot;

[0080] An initial environmental resistance coefficient model is constructed based on a lightweight neural network model, using discretized labels of average slope and ground material type as inputs and the environmental resistance coefficient as output. Preliminary training is conducted using a historical inspection information database, employing a mean squared error loss function to minimize the gap between predicted and actual environmental resistance coefficients. The training batch size and iteration rounds are set based on the training fit. After each inspection plan is completed, newly generated valid data samples are automatically added to the training queue. An adaptive moment estimation online learning algorithm is used to incrementally update the weight parameters of the environmental resistance coefficient model, with a learning rate set to 0.001. Simultaneously, a backup of the parameters before the environmental resistance coefficient model update is retained. If the model's prediction error increases by more than 10% after the update, it is rolled back to the historical version.

[0081] To address the challenges of encountering new terrains for the first time, a meta-learning framework is introduced. By performing meta-training on various known terrains and employing a model-independent meta-learning algorithm, the environmental drag coefficient model obtains initial parameters with generalization capabilities. When faced with new terrains where the ground material type discretization labels are not covered, only a small number of new samples are used for gradient descent updates before the model can be applied to the new task. Furthermore, the new ground material types and their corresponding labels are added to the database to enrich subsequent training data and avoid repeated cold start problems.

[0082] Based on the latest updated environmental resistance coefficient model, the corresponding environmental resistance coefficient is determined by discretizing the labels according to the average slope and ground material type of each connection path.

[0083] Furthermore, setting safe power constraint targets includes:

[0084] Using the robot's current position and each target station as nodes, construct a fully connected graph and generate a set of worst-case paths and direct-connection path distances;

[0085] Due to the high degree of uncertainty in the environment, such as sudden obstacles and terrain changes, the actual path traveled by the robot often deviates from the ideal optimal and shortest path, resulting in actual power consumption far exceeding the estimate. Therefore, the environmental resistance coefficient of the direct path between each target station and the next target station in the worst-case path is obtained. The distance of each direct path is obtained from the set of direct path distances. The estimated travel energy consumption is obtained by multiplying each direct path distance and the environmental resistance coefficient, summing the products, and multiplying them by the robot's energy consumption per unit distance. The expression is as follows:

[0086]

[0087] in, To estimate driving energy consumption, Energy consumption per unit distance for the robot This represents the total number of target sites in the inspection plan. Let be the estimated travel distance of the i-th segment of the path, corresponding to the direct path to each target station; Let be the environmental resistance coefficient of the i-th path segment. The environmental resistance coefficient is dynamically updated based on historical inspection information to achieve adaptive learning of terrain changes and robot state changes.

[0088] Based on the average energy consumption of a single robot stay at a station obtained from the robot information, the estimated stay energy consumption is obtained by calculating the product of the number of selected target stations and the energy consumption of a single robot stay at a station.

[0089] The estimated total energy consumption is obtained by summing the estimated driving energy consumption and the estimated stationary energy consumption.

[0090] A preset safety buffer power level is set to cope with uncertainties, such as detours caused by sudden obstacles, sensor errors, and fluctuations in battery performance. The safety power level is calculated by summing the estimated total energy consumption and the safety buffer power level. After the safety power level is calculated, it is simultaneously sent to the manual control terminal for display, so that the operator can confirm or adjust it manually.

[0091] Furthermore, the set of worst-case path and direct path distances is generated as follows:

[0092] S310: Obtain the estimated driving distance between each target station from the station information, and use it as the weight of the edge between each pair of nodes in the fully connected graph. The greater the distance between the stations, the greater the weight. Use a path planning algorithm, such as Dijkstra's algorithm, to obtain the shortest path D[i][j] between each pair of nodes, and record the shortest distance from node i to node j.

[0093] S320: Using a greedy strategy, starting from the robot's current position S0, select the target station Sf1 that is furthest from the robot's current position, and move the robot from its current position S0 to the target station Sf1. f1 The distance D[S0][S f1 Add the direct path distance to the set and mark the target site S. f1 Sites already visited;

[0094] S330: Among the target sites that have never been visited, find the one furthest from the target site S. f1 The farthest station S f2 Target site S f1 To the target site S f2 Distance D[S f1 ][S f2 Add the direct path distance to the set and mark the target site S. f2 Sites already visited;

[0095] S340: Repeat step S330 until all target sites are visited, and obtain the worst-case path and direct path distance set.

[0096] Furthermore, combining the path planning model, the generated driving path, constrained by safe battery power, includes:

[0097] Select one of the preset uplink / downlink mode, optimal path mode, and free setting mode as the robot's path planning mode;

[0098] When set to uplink / downlink mode, the system obtains the station number or physical coordinates based on the station information, presets the uplink or downlink sequence for the target stations in the inspection plan, and generates a fixed travel path. Based on the fixed travel path, the system performs a safety power constraint check to ensure that the estimated total energy consumption generated by the current path planning does not exceed the safety power. If the check passes, the robot's travel path is determined; otherwise, the system prompts the user to switch the path planning mode. The estimated travel energy consumption is calculated based on the estimated travel distance of each directly connected path in the fixed path, the environmental resistance coefficient, the robot's energy consumption per unit distance, the robot's energy consumption per station, and the number of target stations.

[0099] When set to the optimal path mode, multi-dimensional costs are calculated to achieve multi-objective optimization of safety, inspection efficiency and battery life. A minimum cost path planning model based on the A* algorithm is constructed to obtain the optimal path, and safety power constraints are integrated to determine the robot's travel path.

[0100] When set to free mode, the inspection path is manually set. During each manual setting step, the estimated power consumption is displayed in real time, and an over-limit warning is provided. The environmental map built in real time based on SLAM verifies whether the current inspection path is passable. If passable, it prompts the user to proceed to the next setting step; otherwise, it recommends the optimal path as the free inspection path. When the path planning constraints are met, the robot's travel path is determined.

[0101] Furthermore, the steps for calculating multi-dimensional costs include:

[0102] Calculate site priority by calculating the difference between the preset highest priority and the site priority to obtain the site priority cost;

[0103] Based on robot information, the total battery level and remaining battery level of the robot are extracted and the difference is calculated to obtain the battery consumption level. The higher the battery consumption level, the greater the possibility of the robot running out of power. Based on station information and charging point information, the charging travel distance between each station and the charging point is obtained. By calculating the ratio of the charging travel distance to the farthest distance that the robot can travel when fully charged, the charging convenience level is obtained. The farther the distance between the station and the charging point, the higher the charging convenience index, and the greater the difficulty for the robot to reach the charging point. By calculating the product of the battery consumption level and the charging convenience index, the charging demand cost is obtained.

[0104] Retrieve all records of the distance between the current site and the next target site, and take the average value as the distance cost.

[0105] Furthermore, calculating site priorities includes:

[0106] Based on historical inspection information, the frequency of fault occurrence within a preset period and the frequency of direct path replanning within a preset period are obtained for each site. A higher frequency of fault occurrence indicates a greater risk level at the site and a higher demand for inspection. A higher frequency of direct path replanning indicates a lower accessibility at the site and a higher demand for arranging inspections as soon as possible. The preset period is adjusted according to the scenario to match the pattern of equipment fault occurrence, so as to ensure an accurate reflection of the site status.

[0107] The site priority is obtained by weighted summing the frequency of fault occurrence within a preset period and the frequency of direct path replanning within the preset period, as shown in the following expression:

[0108]

[0109] Wherein, P is the site priority, F is the fault occurrence frequency within the preset period, R is the direct path replanning frequency within the preset period, α is the preset fault frequency weight coefficient, β is the preset direct path replanning frequency weight coefficient, and α+β=1, which is adjusted according to actual inspection needs; specifically, the fault frequency weight coefficient is increased in high fault scenarios, and the direct path replanning frequency weight coefficient is increased in complex terrain scenarios.

[0110] Furthermore, a minimum cost path planning model based on the A* algorithm is constructed to obtain the optimal path, and safety power constraints are integrated, including:

[0111] S410: Normalize and sum the site priority cost, charging demand cost, and distance cost of the sites visited on the current path to calculate the actual cost already traveled; calculate the current estimated total energy consumption based on the distance traveled on the current path, the environmental drag coefficient, and the robot's energy consumption per unit distance.

[0112] S420: Perform a safety power constraint check. If the current estimated total energy consumption exceeds the safety power, it means that there is no feasible direction to extend the current path. Abandon the last station of the current path and backtrack to step S410. If the check passes, proceed to step S430.

[0113] S430: Using each uninspected station as the next target station, calculate the priority cost and charging cost of each uninspected station, as well as the distance cost to other uninspected stations. After normalization, perform equal-weight summation to obtain the remaining cost with each uninspected station as the intermediate node.

[0114] S440: Calculate the total estimated cost by summing the actual cost already incurred and the remaining cost; perform a safety power constraint check, and prioritize the sites that meet the safety power constraint and always minimize the estimated cost, adding them to the visited site sequence.

[0115] S450: Repeat steps S410-S440 until all target stations are traversed, and finally determine the driving path with the lowest total cost from all feasible paths that meet the safe power constraints.

[0116] Furthermore, adaptive adjustment of the robot's motion state includes:

[0117] A preset start threshold (such as 60% of the safe battery level) is set to determine whether the robot's remaining battery level exceeds the preset start threshold. If it does, the robot is controlled to travel along the predetermined path. If it does not exceed the threshold, the robot is controlled to go to the charging point or send a low battery warning to the human control terminal.

[0118] During robot inspection, the estimated total energy consumption of the current inspection plan is dynamically calculated based on the real-time updated environmental drag coefficient, and the operable power threshold is adjusted synchronously. The initial default value of the operable power threshold is set to 40% of the safe power level, and it is adjusted proportionally as the inspection plan progresses. When the robot's remaining power is lower than the real-time adjusted operable power threshold, the current inspection task is immediately stopped, the GPS / SLAM coordinates of the power failure location are recorded, and the robot is controlled to go to the charging station until the robot's remaining power is higher than the operable power threshold. Then, the robot returns to the power failure location and continues the inspection.

[0119] Example 2

[0120] refer to Figure 4 As shown in the figure, this embodiment introduces a robot remote control inspection system, including: a data acquisition module, a data storage module, a planning and control module, a communication module, and a display module;

[0121] The data acquisition module includes:

[0122] The positioning unit is used to acquire the robot's physical coordinates in real time;

[0123] A multimodal sensor integration unit is used to acquire robot information, ground slope, and inspection data;

[0124] The data storage module is used to store the information acquired by the data acquisition module, including site information, charging point information, historical inspection information, and intelligent learning model parameters.

[0125] The planning and control module includes:

[0126] The intelligent computing unit is used to calculate the estimated power consumption, site priority, multi-dimensional costs, and plan the driving route, and dynamically update the environmental resistance coefficient prediction model based on historical inspection data.

[0127] The data conversion unit is used to convert the data from the computing unit into executable motion commands and to convert manual control signals into drive signals.

[0128] The configuration unit is used to switch control permissions between the management system and the manual remote control terminal, and to set the robot's driving mode.

[0129] The communication module is used for real-time transmission of data and instructions;

[0130] The display module is used to display information and receive manual operations, providing functions such as setting inspection plans, switching permissions and driving modes, planning routes, controlling the robot, and displaying alarm information.

[0131] In summary, this invention, by constructing a safety power constraint model based on worst-case paths, fundamentally ensures the completion of inspection tasks in complex environments and improves operational reliability. By combining a meta-learning mechanism to adaptively update the environmental resistance coefficient, it achieves high-precision energy consumption prediction for dynamically changing environments. Through path planning using the A* algorithm that integrates multi-dimensional cost and safety power constraints, it intelligently balances inspection efficiency, the urgency of risk handling, and endurance safety while ensuring safety.

[0132] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for remote-controlled inspection of a robot, characterized in that: S1: In response to manual remote control commands, construct an inspection plan based on one or more target stations selected by the manual remote control terminal; S2: Acquire robot information, site information, charging point information, and historical inspection information, and adaptively update the environmental resistance coefficient based on the meta-learning mechanism, and set a safe power constraint target; S3: Combines path planning mode to dynamically generate driving paths constrained by safe battery level; S4: Set the robot driving mode, and after controlling the robot's movement, adaptively adjust the robot's movement state based on the dynamic comparison result between the robot's remaining power and the safe power constraint; S5: Real-time acquisition and storage of robot information, site information, inspection data and alarm information until the inspection plan is completed and robot control is handed over.

2. The robot remote-controlled inspection method according to claim 1, characterized in that, The adaptive update of the environmental resistance coefficient based on the meta-learning mechanism includes: The driving data corresponding to the connection path between each target station and the next numbered target station in each inspection task is obtained. After data cleaning, a historical inspection information database is constructed. The driving data includes discretized labels of the average slope and ground material type corresponding to each connection path. Based on the physical model, the actual environmental resistance coefficient is calculated by inversely calculating the ratio of the actual driving power consumption to the product of the connection path length and the robot's energy consumption per unit distance. A lightweight neural network model is constructed, with discretized labels of average slope and ground material type as input and environmental resistance coefficient as output. It is initially trained based on a historical inspection information database to minimize the difference between the predicted value and the actual environmental resistance coefficient, thus obtaining an initial environmental resistance coefficient model. Newly generated valid data samples are automatically added to the training queue for incremental updates; a meta-learning mechanism is introduced to quickly generate an initial prediction model for new terrain to overcome the cold start problem. Based on the latest updated environmental resistance coefficient model, the corresponding environmental resistance coefficient is determined by discretizing the labels according to the average slope and ground material type of each connection path.

3. The robot remote-controlled inspection method according to claim 1, characterized in that, Setting safe power constraint targets includes: Using the robot's current position and each target station as nodes, construct a fully connected graph and generate a set of worst-case paths and direct-connection path distances; Obtain the estimated driving distance between each target station and the next numbered target station from the station information; The environmental drag coefficients and distances of the direct paths between each target station and the next target station in the worst-case path are obtained. The estimated driving energy consumption is obtained by calculating the product and summing them. The average energy consumption of the robot at a single station is obtained. The estimated energy consumption at a single station is obtained by multiplying the number of selected target stations by the energy consumption of the robot at a single station. The estimated total energy consumption is obtained by summing the estimated driving energy consumption and the estimated stationary energy consumption. A preset safety buffer power level is set, with the total estimated energy consumption plus the safety buffer power level being the safe power level.

4. The robot remote-controlled inspection method according to claim 3, characterized in that, The set of worst-case path and direct path distances includes: S310: Obtain the estimated driving distance between each target station from the station information, and use it as the weight of the edge between each pair of nodes in the fully connected graph; obtain the shortest path between each pair of nodes through the path planning algorithm; S320: Using a greedy strategy, starting from the robot's current position S0, select the target station Sf1 that is furthest from the robot's current position, and move the robot from its current position S0 to the target station Sf1. f1 The distance D[S0][S f1 Add the direct path distance to the set and mark the target site S. f1 Sites already visited; S330: Among the target sites that have never been visited, find the one furthest from the target site S. f1 The farthest station S f2 Target site S f1 To the target site S f2 Distance D[S f1 ][S f2 Add the direct path distance to the set and mark the target site S. f2 Sites already visited; S340: Repeat step S330 until all target sites are visited, and obtain the worst-case path and direct path distance set.

5. The robot remote-controlled inspection method according to claim 1, characterized in that, Combining the path planning model, the generated driving path, constrained by safe battery power, includes: Select one of the preset path planning modes as the robot's path planning mode; When set to uplink / downlink mode, the station number or physical coordinates are obtained based on the station information. The uplink or downlink sequence of the target station in the inspection plan is preset. When the path planning constraints are met, the robot's travel path is determined. When set to optimal path mode, the robot's travel path is determined by calculating multi-dimensional costs, constructing a minimum cost path planning model based on the A* algorithm to obtain the optimal path, and integrating safety power constraints. When set to free mode, the inspection path is manually set and a feasibility check is performed at each step. When the check passes and the path planning constraints are met, the robot's travel path is determined.

6. The robot remote-controlled inspection method according to claim 5, characterized in that, Calculating multi-dimensional costs includes: Calculate site priority by calculating the difference between the preset highest priority and the site priority to obtain the site priority cost; Based on the robot information, the total battery power and remaining battery power of the robot are extracted and the difference is calculated to obtain the degree of battery consumption; based on the station information and charging point information, the charging travel distance between each station and the charging point is obtained, and the charging convenience is obtained by calculating the ratio of the charging travel distance to the farthest distance that the robot can travel when fully charged; The cost of charging demand is obtained by multiplying the power consumption level and the charging convenience index. Retrieve all records of the distance between the current site and the next target site, and take the average value as the distance cost.

7. A robot remote-controlled inspection method according to claim 6, characterized in that, Calculating site priorities includes: Based on the historical inspection information, obtain the frequency of fault occurrence within a preset period of the site and the frequency of replanning of direct connection paths within a preset period of the path. The site priority is obtained by weighted summing of the failure frequency within a preset period and the direct path replanning frequency within the preset period.

8. A robot remote-controlled inspection method according to claim 5, characterized in that, A minimum cost path planning model based on the A* algorithm is constructed to obtain the optimal path, and safety power constraints are integrated, including: S410: Normalize and sum the site priority cost, charging demand cost, and distance cost of the sites that have been inspected, and calculate the actual cost already traveled. S420: Perform a safety power constraint verification. If the verification fails, backtrack to step S410; if the verification passes, proceed to step S430. S430: For uninspected stations, calculate their own priority and charging cost, as well as the distance cost to other uninspected stations, and sum them up to obtain the remaining cost. S440: Calculate the total estimated cost by summing the actual cost already incurred and the remaining cost; perform a safety power constraint check, and prioritize the site that meets the safety power constraint and always minimizes the estimated cost. S450: Repeat steps S410-S440 until all target stations are traversed, and finally determine the driving path with the lowest total cost from all feasible paths that meet the safe power constraints.

9. A robot remote-controlled inspection method according to claim 1, characterized in that, Adaptive adjustment of robot motion state includes: The robot determines whether its remaining battery power exceeds the preset start threshold. If it does, the robot is controlled to travel along a predetermined path. Otherwise, it goes to the charging station or issues a low battery warning. During the robot inspection process, the estimated total energy consumption of the current inspection plan is dynamically calculated, and the operable power threshold is adjusted simultaneously; based on the comparison results of the robot's remaining power and operable power threshold obtained in real time, the robot's motion state is adjusted.

10. A robot remote-controlled inspection system, used to implement a robot remote-controlled inspection method as described in any one of claims 1-9, characterized in that, include: Data acquisition module, data storage module, planning and control module, communication module, display module; The data acquisition module includes: The positioning unit is used to acquire the robot's physical coordinates in real time; A multimodal sensor integration unit is used to acquire robot information, ground slope, and inspection data; The data storage module is used to store the information acquired by the data acquisition module, including site information, charging point information, historical inspection information, and intelligent learning model parameters. The planning control module includes: The intelligent computing unit is used to calculate the estimated power consumption, site priority, multi-dimensional costs, and plan the driving route, and dynamically update the environmental resistance coefficient prediction model based on historical inspection data. The data conversion unit is used to convert the data from the computing unit into executable motion commands and to convert manual control signals into drive signals. The configuration unit is used to switch control permissions between the management system and the manual remote control terminal, and to set the robot's driving mode. The communication module is used for real-time transmission of data and instructions; The display module is used to display information and receive manual operations, providing functions such as setting inspection plans, switching permissions and driving modes, planning routes, controlling robots, and displaying alarm information.

Images