A low-power inspection task processing system for an inspection robot

Abstract

The application discloses a kind of low power inspection task processing systems of inspection robot, including central cooperative processing server and at least two inspection robots, both interactive connection, inspection robot is equipped with vehicle control unit, vehicle control unit real-time monitoring battery remaining power, preset first node, when the first time that electric quantity drops to the node, trigger low power early warning signal, central cooperative processing server contains state aggregation perception module, dynamic energy consumption prediction module and task reconstruction decision module, the former obtains aggregated robot real-time state data, dynamic energy consumption prediction module predicts task energy consumption, after receiving early warning, task reconstruction decision module selects execution task replacement agreement or task exchange agreement according to real-time state data and predicted energy consumption, to guarantee inspection task continuity, data integrity and system operation robustness.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of train inspection, more particularly to a low-battery inspection task processing system for an inspection robot. BACKGROUND

[0002] With the increasing demand for railway train inspection, the application of inspection robots in automated detection and intelligent management has gradually become a research hotspot. However, in actual inspection tasks, the inspection robot may not be able to complete the entire inspection task due to insufficient power, especially in complex inspection environments or when data collection takes a long time. Therefore, how to effectively handle tasks in the low-battery state of the inspection robot has become a technical problem to be solved.

[0003] After searching, a train inspection robot control system with publication number CN116476099B is disclosed. This patent realizes efficient detection of train bottom parts by means of real-time collection of road information, generation of positioning data, and identification of obstacles. However, this technical solution does not involve task allocation and processing mechanisms for inspection robots in low-battery states, which may cause the robot to interrupt the inspection due to insufficient power when performing complex or time-consuming tasks, affecting the overall task completion efficiency. In addition, this solution lacks dynamic adjustment of inspection task priorities and cannot reasonably plan inspection paths or task decomposition based on power states, which may increase the risk of task failure.

[0004] After searching, a transfer device and transfer method for a train inspection robot with publication number CN116986232B are disclosed. This patent designs a flexible transfer device to enable the inspection robot to be quickly transferred to different locations, eliminating the limitations of infrastructure reconstruction and fixed transfer locations. However, this technical solution mainly focuses on the transfer of the inspection robot before or after the task starts, and does not fully consider the situation where the robot needs emergency handling due to insufficient power during task execution. For example, when the robot runs out of power during inspection, this solution cannot provide effective emergency measures or task handover mechanisms, which may cause the inspection task to be forced to interrupt or delayed. In addition, this solution does not mention task reassignment strategies when multiple robots work together, making it difficult to coordinate the overall task when a single robot runs out of power.

[0005] The above problems show that the existing inspection robot technology still has certain deficiencies in the task processing capability in the low-battery state, especially in terms of task dynamic adjustment, path optimization, collaborative work, and emergency transfer, which lack systematic solutions. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application aims to overcome the defects of task interruption, data loss and low efficiency of multi-robot cooperation in low power working conditions caused by the contradiction between static task planning and dynamic energy consumption in the prior art task execution system of the inspection robot.

[0007] Therefore, the present application provides a low power inspection task processing system of an inspection robot, which realizes accurate quantification, prediction and management of the energy state of a robot cluster by constructing a closed-loop control system with a predicted energy model as the core, and on this basis, establishes a multi-threshold triggered collaborative processing protocol containing task dynamic reconstruction, intelligent replacement and peer exchange, thereby maximizing the continuity of the inspection task, the integrity of the data and the robustness of the overall operation of the system while ensuring the safe return of the individual robots.

[0008] To achieve the above object, the present application provides the following technical scheme:

[0009] A low power inspection task processing system of an inspection robot, comprising a central collaborative processing server and at least two inspection robots, the central collaborative processing server being interactively connected with the inspection robots, and the inspection robots being provided with a vehicle-mounted control unit, characterized in that the vehicle-mounted control unit monitors the remaining power of the battery carried by itself in real time and is provided with a first node, and when the remaining power is first detected to drop to the first node, a low power warning signal is triggered.

[0010] The central collaborative processing server comprises:

[0011] A state aggregation perception module that acquires and aggregates real-time state data reported by all inspection robots, the real-time state data at least including the current position coordinates and the remaining power of each robot;

[0012] A dynamic energy consumption prediction module that predicts the energy consumption required for executing the task based on a preset dynamic energy consumption prediction model and the real-time state information and the to-be-executed task;

[0013] A task reconstruction decision module that selects an execution collaborative processing protocol according to the real-time state data and the predicted energy consumption after obtaining the low power warning signal, the collaborative processing protocol comprising:

[0014] A task replacement protocol that selects an optimal replacement robot from other inspection robots in a non-low power state and instructively assigns the uncompleted task of the inspection robot triggering the low power warning signal to the replacement robot for execution;

[0015] Task exchange protocol: the inspection robot triggering the low power warning signal and another non-low power state inspection robot complete the equal exchange of the remaining task package of the two under the condition of meeting the preset energy exchange feasibility and global time efficiency check.

[0016] Further, the task replacement protocol includes a selection strategy, which includes defining the inspection robot triggering the low power warning signal as the first robot, after receiving the task package to be handed over sent by the first robot, starting the optimal replacement robot screening program, the screening program traverses all other candidate robots in non-low power state, and quantitatively evaluates each candidate robot based on a multi-dimensional task affinity score model, and finally selects the candidate robot with the highest task affinity score as the optimal replacement robot.

[0017] Further, the dimension factors in the multi-dimensional task affinity score model include spatial proximity factor, energy adequacy factor, path coordination index factor, historical task quality score factor and current task load factor, the spatial proximity factor reflects the inverse of the Euclidean distance from the current position of the candidate robot to the first task point in the task package to be handed over, the energy adequacy factor reflects the difference between the current remaining power of the candidate robot and the estimated total energy consumption required to complete the task package to be handed over, the path coordination index factor reflects the increase in the total path length after inserting the task package to be handed over into the existing task sequence of the candidate robot, the historical task quality score factor reflects the comprehensive index of the data quality of the past completed task points of the candidate robot, and the current task load factor reflects the number of remaining tasks or the estimated remaining working time of the candidate robot.

[0018] Further, the vehicle-mounted control unit further includes a preset dynamic energy consumption prediction model, which performs reachability evaluation on all task points in its remaining task list one by one, the reachability reflects the total estimated energy consumption required to complete the remaining task points one by one from the current position, and compares it with a preset available operating energy, the available operating energy reflects the difference between the first node and a second node lower than the first node, and deducts the return safety margin dynamically calculated according to the distance from the current position to the charging station, if the evaluation result is that the available operating energy meets the completion of at least one task point in the remaining task list, determines the maximum number of task point sets that can be completed before the power is exhausted to the second node as the final execution sequence, and packs the remaining task points that cannot be completed into a task package to be handed over and sends it to the central collaborative processing server.

[0019] Further, the task exchange protocol includes a preset check condition, the preset check condition includes:

[0020] The energy consumption interchange feasibility check, the current remaining power of the inspection robot triggering the low power warning signal is greater than the estimated energy consumption required for completing the remaining task package of another non-low power state inspection robot;

[0021] The global time efficiency check, the sum of the time consumed by the inspection robot triggering the low power warning signal and another non-low power state inspection robot moving from the current position to the task starting point of the other party is less than or equal to the preset time window threshold;

[0022] The sensor capability compatibility check, the sensor configuration carried by the inspection robot triggering the low power warning signal and the other non-low power state inspection robot respectively needs to be able to meet the data acquisition of all task points in the task package of the other party.

[0023] Further, when it is determined that there is an exchange object that meets all the check conditions, the central cooperative processing server simultaneously issues a task exchange instruction to the inspection robot triggering the low power warning signal and the exchange object, the task exchange instruction includes the remaining task package information and the optimal path planning to the task starting point of the other party, and if there are differences in the same type of sensor model or calibration parameters between the inspection robot triggering the low power warning signal and the exchange object, the task exchange instruction also includes a data acquisition parameter conversion matrix.

[0024] Further, the central cooperative processing server further includes an initial task allocation module,

[0025] The initial task allocation module clusters the three-dimensional coordinates of a plurality of to-be-executed task points based on a density-based spatial clustering algorithm to form a plurality of task point clusters that are adjacent in space, calculates a total task amount for each task point cluster, the total task amount reflects the sum of the baseline time estimation value of all task points in the cluster and the estimated moving time of the shortest traversal path between the task points in the cluster, and then distributes the task point clusters to different inspection robots according to an optimization allocation algorithm to minimize the variance of the predicted remaining power of all robots after completing their respective tasks, thereby forming an initial task list for each robot.

[0026] Further, the central cooperative processing server further includes a data fusion and report generation module,

[0027] The data fusion and report generation module accurately associates all data collected by different inspection robots at different times and through different cooperative processing protocols with the originally defined key detection points according to the globally unique task point identifier attached to each data file, and generates a comprehensive inspection report.

[0028] Further, the input quantity of the dynamic energy consumption prediction model includes inherent physical parameters of the inspection robot, real-time kinematic parameters, real-time load parameters, real-time power consumption of a sensor system and real-time computing load of a vehicle-mounted computing unit.

[0029] Further, the dynamic energy consumption prediction module calculates the instantaneous total power consumption of the robot in real time by weighted summation according to driving power, sensor system power and computing unit power, wherein the driving power reflects a function of real-time kinematic parameters and real-time load parameters, the sensor system power is determined according to an activated sensor list and working states thereof, and the computing unit power reflects a real-time load rate of the vehicle-mounted computing unit.

[0030] The present application has the following beneficial effects: 1. The vehicle-mounted control unit monitors the electric quantity in real time and presets a low-electricity warning node, after triggering the warning, the task reconstruction decision module of the central cooperative processing server reasonably distributes the unfinished tasks of the low-electricity robot to other robots through a task replacement or task exchange protocol, effectively solving the problem of task interruption and data loss of the inspection robot caused by insufficient electric quantity in the prior art, ensuring the continuity of the inspection task, in addition, the dynamic energy consumption prediction model considers multiple factors such as inherent physical parameters of the robot, real-time kinematic parameters, load parameters, sensor power consumption and computing load, calculates the instantaneous total power consumption in real time through weighted summation, and continuously optimizes the model combined with real-time data, greatly improving the accuracy of energy consumption estimation, providing a reliable quantitative basis for task replacement, exchange and reachability evaluation;

[0031] 2. The task replacement protocol in the present application filters the optimal replacement through a multi-dimensional task affinity scoring model, covering spatial proximity, energy abundance, path coordination, historical task quality and current load, ensuring efficient execution of the replaced task and reducing resource waste, while the task exchange protocol realizes equal exchange of tasks when the conditions are met through three verifications of energy consumption feasibility, time efficiency and sensor compatibility, fully utilizes the remaining electric quantity and capacity of each robot, improves the overall inspection efficiency, in addition, the reachability evaluation mechanism of the vehicle-mounted control unit determines the maximum task set that can be completed by comparing the available operating energy and the remaining task energy consumption when the electric quantity is low, and reserves a safe return margin, ensuring that the robot can safely return while maximizing the completion of tasks, achieving a balance between efficiency and safety. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is the overall flowchart in the present application;

[0033] Figure 2 is the task replacement protocol execution flowchart in the present application;

[0034] Figure 3 is the task exchange protocol execution flowchart in the present application. Detailed Implementation

[0035] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0036] Because existing inspection robot technology still has certain shortcomings in terms of task processing capabilities under low battery conditions, especially in the areas of dynamic task adjustment, path optimization, collaborative operation, and emergency transfer, this invention designs a low-battery inspection task processing system for inspection robots, such as... Figure 1 As shown, it comprises a central collaborative processing server and at least two inspection robots. The central collaborative processing server is interactively connected to the inspection robots, and the inspection robots are equipped with vehicle-mounted control units.

[0037] The central collaborative processing server integrates a task definition and decomposition module, a system state initialization module, a dynamic energy consumption prediction module, an initial task allocation module, a state aggregation and perception module, a task reconstruction decision module, and a data fusion and report generation module. The task definition and decomposition module receives and parses externally input train maintenance instructions. These instructions include the unique identifier of the train to be inspected, its model data, its precise three-dimensional spatial parking posture information on the maintenance track, and a structured list of key inspection points. This module parses and transforms each point in this list into a task point. Each task point is assigned a globally unique identifier and contains a set of parameters necessary to execute the task unit. Specifically, this parameter set includes the three-dimensional spatial coordinates of the target point. In conjunction with sensor modalities, such as high-resolution visual image acquisition, 3D laser point cloud scanning, or ultrasonic thickness measurement, as well as sensor parameter configuration, a specific task point data structure specification can be defined as: {AIU_ID:1001,TargetCoord:{x:45.72m,y:1.25m,z:0.88m},SensorModality:'HighResVisual',SensorParams:{ExposureTime:1500μs,Gain:18dB,Focus:'auto'},EstimatedDuration: 2.8s,Priority: 1}.

[0038] The system state initialization module is responsible for broadcasting a state query instruction to all standby inspection robots at the beginning of a task. Each robot's on-board control unit returns its initial state vector, including the robot's current accurate position coordinates, current remaining battery percentage, and robot hardware health status code, upon receiving the instruction. The central server collects all initial state vectors of the robots to complete the initial inventory of system resources.

[0039] The dynamic energy consumption prediction module establishes and maintains an independent and continuously iterative dynamic energy consumption prediction model for each inspection robot on the server side. This model is not a static lookup table or linear extrapolation model, but a composite function that takes into account multiple physical and computational factors. The input variables of this model include the robot's inherent physical parameters such as overall mass, drive motor rated power and efficiency curve, wheel structure and transmission ratio; real-time kinematic parameters of the robot such as current linear speed, angular speed and acceleration; real-time feedback of drive motor load current from the on-board control unit, which is used to inversely calculate the rolling friction coefficient of the robot on the current road surface; real-time power consumption of the sensor system, which is dynamically calculated based on the currently activated sensor type and its working mode; and real-time load rate of the on-board control unit to quantify the computational energy consumption generated by path planning, obstacle avoidance, etc. The model calculates the instantaneous total power consumption of the robot in real time through a weighted summation formula , specifically as follows: , is the drive power, which is a complex function of speed, acceleration, and motor current, is the sensor power, which is obtained by looking up the activated sensor list and its working state, is the computational power consumption, which is linearly related to the load rate of CPU and GPU, , , is the weight coefficient used to balance the contribution of each part of the energy consumption, and is an overall calibration coefficient, which is calibrated by performing a complete discharge test on the robot under standard working conditions, and its value is usually between 1.0 and 1.1 to compensate for unmodeled factors such as battery aging and internal resistance changes. By time integrating the instantaneous total power along a planned path or within an execution cycle at a task point, the system can accurately predict the energy consumption required to execute any task segment. During task execution, the model uses real-time power consumption data reported by the robot to continuously update and optimize the model's internal parameters, particularly the rolling friction coefficient and the calibration coefficient , through the Kalman filtering algorithm, thereby achieving adaptive iteration of the model.

[0040] The initial task allocation module, based on the spatial distribution characteristics of all task points and the initial state of each robot, performs the first task division. First, the module calls a density-based spatial clustering algorithm, such as the DBSCAN algorithm, to cluster the three-dimensional coordinates of all task points to form several spatially adjacent task point clusters. Then, the system calculates the total amount of tasks for each cluster, which is the sum of the baseline time estimation value of all task points in the cluster and the estimated moving time of the shortest traversal path between the task points in the cluster. Finally, through a distribution strategy that maximizes global energy efficiency, such as genetic algorithm optimization distribution, the task point clusters are distributed to different robots to form their initial task lists, with the goal of making the estimated remaining power of all robots after completing their respective tasks as balanced as possible.

[0041] The state aggregation perception module continuously receives real-time state update data packets from each robot's on-board control unit during task execution. The data packets are sent at a fixed high frequency, and their contents constitute a complete snapshot of the robot's current state, including high-precision real-time position coordinates, instantaneous speed, accurate remaining power, currently executed task point identifier, completed task point list, remaining task point list, real-time current of drive motor, and load of on-board computing unit. The central collaborative server aggregates, analyzes, and timestamps aligns these data streams from different robots, builds and maintains a global "collaborative situation map" in a shared memory database. This map not only visually presents the position trajectories, power decline curves, and task progress of all robots, but also provides continuous and real input data for subsequent dynamic energy consumption prediction models and decision-making basis for multi-threshold task reconstruction decision-making module.

[0042] The on-board control unit of the inspection robot is used to execute instructions from the central server and real-time feedback of its own state. The on-board control unit is configured with modules to read real-time voltage, current, and remaining capacity data of the battery. The on-board control unit is also configured with local power thresholds: first node and second node. In this invention, the first node is set to 20%, and the second node is set to 10%.

[0043] When the on-board control unit of an inspection robot detects that its own power has first dropped and touched the first node, the task reconstruction decision-making module is activated, and according to the current state of the robot and the global collaborative situation, one of the following two logic branches is selected to define the inspection robot that triggers the low power warning signal as the first robot.

[0044] As Figure 2As shown, the task handover protocol is triggered when the first robot reaches the first node, and its on-board control unit invokes the local dynamic energy consumption prediction model to perform a one-by-one reachability assessment of all task points in its remaining task list. The assessment method is to calculate the total estimated energy consumption required to complete the sequence of remaining task points from the current location, including movement energy consumption and operation energy consumption, and compare it with the current available remaining energy, which is the difference between the power from the first node to the second node, minus the reserved return safety margin. If the assessment result shows that the available remaining energy of the first robot is sufficient to complete at least one task point in its remaining task list, the protocol is started, and the specific execution process is as follows: the on-board control unit of the first robot first determines the maximum number of task point subsets that it can complete before running out of power to the second order, and marks this subset as the final execution sequence. At the same time, it packages the remaining and uncompleted task points into a handover task package, which contains all the detailed parameters of these task points, and sends it to the central collaborative processing server through the wireless communication link. The first robot then continues to execute its final execution sequence.

[0045] At the same time, the central server starts the optimal handover robot screening program immediately after receiving the handover task package. This program iterates through all other robots currently executing tasks, excluding the first robot and other robots already in the return or low power state, and performs a quantitative assessment of each candidate robot, defined as the second robot. The assessment algorithm is based on a multi-dimensional task affinity score model. The calculation formula of the model is . Wherein, is the Euclidean distance from the current position of the second robot to the first task point in the handover task package, representing spatial proximity; is the current remaining power of the second robot, is the estimated total energy consumption required to complete the task package, and the difference between the two represents the energy sufficiency; is a path coordination index that quantifies the increase in total path length after inserting the handover task into the second robot's existing task sequence. The smaller the increase, the higher the coordination; is the historical task quality score of the second robot, a comprehensive index based on the data quality of the task points it has completed in the past (such as image clarity, point cloud completeness); is the current remaining task load of the second robot; , , , and are preset and can be adjusted by the system administrator, the server calculates the scores of all candidate robots, and selects the robot with the highest score as the final replacement; then, the server pushes the task package to be handed over to the replacement robot, which seamlessly integrates it into its own task queue, plans a new optimal execution path, and executes the replacement task after completing its original task or at an appropriate node in the optimal path.

[0046] In the process of the first robot executing its final execution sequence, if its power further decreases and reaches the second node, its on-board control unit will immediately and unconditionally suspend all current inspection operations, and start the autonomous return program. The return path is planned by its local path planning module using the A algorithm on the pre-installed environment map containing permanent obstacles and dynamic obstacle information issued by the server, to find a shortest and collision-free path back to the charging pile. The coordinate point sequence of the return path is uploaded to the central server in real time. After receiving the return path, the server immediately marks it as a high-priority reserved space-time channel in the global collaborative situation map, and sends path warning and avoidance instructions to other robots that may collide along the path, ensuring the absolute safety of the return process. If there are two or more candidate robots that meet the conditions when screening for replacement robots, in addition to considering the scores, the system will preferentially select the robot with a higher historical task quality score and a more optimal path coordination index.

[0047] As shown in Figure 3 The task exchange protocol, whose trigger condition is: when the first robot reaches the first node, its local reachability evaluation result shows that it has insufficient available remaining energy to complete any of the tasks in its remaining task list. In this case, the first robot broadcasts a task exchange request to the central server, which contains the complete state vector of the first robot and its unfinished remaining task package A.

[0048] Upon receiving the request, the central server initiates a peer exchange matching program, which iterates through all other robots that are not low on power as potential exchange partners, such as the second robot, and conducts a rigorous feasibility check on each possible exchange combination, which includes three necessary conditions: first, an energy exchange feasibility check, i.e., the first robot's current remaining power must be greater than the estimated energy consumption required to complete the second robot's remaining task package B, and the second robot's current remaining power must be greater than the estimated energy consumption required to complete the first robot's remaining task package A; second, a global time efficiency check, i.e., the sum of the time taken by the first robot to move from its current location to the starting point of the second robot's task and the time taken by the second robot to move from its current location to the starting point of the first robot's task must be less than or equal to a pre-set threshold, which is set to ensure that task exchange does not significantly prolong the overall project duration; third, a sensor capability compatibility check, i.e., the sensor configurations carried by the first robot and the second robot must be able to meet the data acquisition requirements of all task points in the other's task package.

[0049] Only when a pair of robots simultaneously satisfy all three conditions above is the exchange deemed feasible. If there are multiple feasible exchange partners, the server will select the combination that results in the smallest total path increment for the two robots after the exchange. Once the exchange partner is determined, the central server sends task exchange instructions to both the first robot and the exchange partner, which, for the first robot, includes the entire information of the exchange partner's remaining task package and the optimal path planning to the starting point of the task package; similarly, the exchange partner's instructions include the first robot's remaining task package A. In addition, if there are differences in sensor models or calibration parameters between the first robot and the exchange partner, the instructions will also include a data acquisition parameter conversion matrix, which is automatically generated by the server based on pre-stored sensor technical specifications, to guide the robots to adjust their sensor settings when executing the exchanged tasks, thereby ensuring the consistency and comparability of the acquired data. Upon receiving the instructions, the first robot and the exchange partner immediately abandon their original remaining tasks and execute the new tasks from the exchange.

[0050] The data fusion and report generation module is responsible for summarizing all uploaded task point data after all inspection tasks are completed (including initially assigned tasks, handed over tasks, and exchanged tasks). The module accurately correlates data collected by different robots at different times back to the original train key detection points through the globally unique task point identifier attached to each data file. The system automatically checks the quality of the data, such as evaluating the sharpness and contrast of images using image processing algorithms and checking the density and integrity of point clouds using point cloud analysis algorithms. Finally, the system generates a complete and seamlessly spliced comprehensive inspection report. In this report, each item of detection data is accompanied by detailed data traceability information, clearly indicating which robot completed the collection under which coordination mode, thereby providing a complete evidence chain for subsequent quality traceability and operation and maintenance decision-making.

[0051] Embodiment:

[0052] An analog 150-meter-long, 20-meter-wide train maintenance depot, in which a model of an 8-carriage EMU is parked, a total of three inspection robots consistent with the specifications described above are deployed, namely R01, R02, and R03. A total of 120 inspection points are defined, distributed on both sides and the bottom of the train. The initial battery capacity of the robots is 100% (40Ah), the first node threshold is set to 20% (8Ah), the second node threshold is set to 10% (4Ah), and the return safety margin is set to 2Ah.

[0053] After the initial task allocation module runs, the three robots are allocated similar task quantities: R01 is allocated 42 task points, R02 is allocated 38 task points, and R03 is allocated 40 task points.

[0054] After the task starts to execute, the three robots work as planned. At about 1 hour and 48 minutes into the task, the on-board control unit of R01 detects that its remaining battery capacity has reached the first node for the first time. At this time, R01 has completed 32 task points and has 10 task points remaining to be executed.

[0055] The local energy consumption model of R01 immediately starts the reachability assessment, and the assessment result shows that it is estimated to consume 1.8Ah of battery capacity to complete the first three of the remaining 10 task points (AIU-R01-33, AIU-R01-34, AIU-R01-35) from the current location, while its available operating energy is 8Ah-4Ah-2Ah (safety margin) = 2Ah. Therefore, R01 determines its final execution sequence as these three task points and sends the remaining seven task points (AIU-R01-36 to AIU-R01-42) packaged into a handover task package to the central server.

[0056] The central server receives the task package and immediately evaluates R02 and R03, and the real-time state at that time is as follows:

[0057] R02: remaining power 22Ah (55%), remaining tasks 8, current position 15 meters from the starting point of the task handover, historical task quality score 92 points.

[0058] R03: remaining power 25Ah (62.5%), remaining tasks 11, current position 48 meters from the starting point of the task handover, historical task quality score 88 points.

[0059] The server calculates that the energy required to complete the 7 task points in the handover package is 6.5Ah.

[0060] Score calculation (weight =0.35, =0.30, =0.15, =0.10, =0.10):

[0061] TAS_R02=0.35*(1 / (1+15))+0.30*(22-6.5)+0.15*(0.92)+0.10*(92 / 100)-0.108=0.022+4.65+0.138+0.092-0.8=4.102;

[0062] TAS_R03=0.35(1 / (1+48))+0.30*(25-6.5)+0.15*(0.85)+0.10*(88 / 100)-0.10*11=0.007+5.55+0.128+0.088-1.1=4.673;

[0063] The calculation result shows that although R02 is closer in space, R03 has more abundant energy and lower current task load, and its score is higher, so the server selects R03 as the successor and issues the task package to R03.

[0064] R01 continues to perform and completes its final 3 task points, and then safely starts the return program when the power drops to the second node, R03 seamlessly connects to perform the 7 task points from R01 after completing its own tasks, and R02 completes all its tasks as planned.

[0065] Finally, all 120 task points are successfully executed, data collection is complete, and the three robots safely return to the charging pile after completing the task. The entire inspection task takes 2 hours and 55 minutes.

[0066] The above are only preferred embodiments of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical scheme falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall also be considered as falling within the protection scope of the present application.

Claims

1. A low-battery inspection task processing system for inspection robots, comprising a central collaborative processing server and at least two inspection robots, wherein the central collaborative processing server is interactively connected to the inspection robots, and the inspection robots are equipped with vehicle-mounted control units, characterized in that: The vehicle control unit monitors the remaining power of its battery in real time and has a preset first node. When the remaining power drops to the first node for the first time, a low power warning signal is triggered. The central collaborative processing server includes: The status aggregation and perception module acquires and aggregates real-time status data reported by all inspection robots. The real-time status data includes at least the current position coordinates and remaining battery power of each robot. The dynamic energy consumption prediction module predicts the energy consumption required to execute the task based on the preset dynamic energy consumption prediction model, using real-time status information and the task to be executed. The task reconfiguration decision module, upon receiving a low battery warning signal, selects to execute a collaborative processing protocol based on real-time status data and predicted energy consumption. The collaborative processing protocol includes: The task succession protocol selects the optimal successor robot from other inspection robots that are not in a low-battery state, and instructively assigns the unfinished tasks of the inspection robot that triggered the low-battery warning signal to the successor robot for execution. Task exchange protocol: The inspection robot that triggers the low battery warning signal and another inspection robot that is not in a low battery state will complete the equal exchange of their remaining task packages under the conditions of satisfying the preset energy exchange feasibility and global time efficiency verification. If the inspection robot that triggered the low battery warning signal has enough remaining energy to complete at least one task in its remaining task list, it executes the task succession protocol; if it has insufficient remaining energy to complete any task in its remaining task list, it executes the task exchange protocol.

2. The low-battery inspection task processing system for an inspection robot according to claim 1, characterized in that: The task succession protocol includes a selection strategy, which includes defining the inspection robot that triggers the low battery warning signal as the first robot. After receiving a handover task package sent by the first robot that it cannot complete, an optimal successor robot screening program is started. The screening program traverses all other candidate robots that are not in a low battery state and performs a quantitative evaluation of each candidate robot based on a multi-dimensional task affinity scoring model. Finally, the candidate robot with the highest task affinity score is selected as the optimal successor robot.

3. The low-battery inspection task processing system for an inspection robot according to claim 2, characterized in that: The multi-dimensional task affinity scoring model includes spatial proximity factor, energy sufficiency factor, path coordination index factor, historical task quality score factor, and current task load factor. The spatial proximity factor reflects the reciprocal of the Euclidean distance from the candidate robot's current position to the first task point in the task package to be handed over. The energy sufficiency factor reflects the difference between the candidate robot's current remaining power and the estimated total energy consumption required to complete the task package to be handed over. The path coordination index factor reflects the increase in the total path length after the task package to be handed over is inserted into the candidate robot's existing task sequence. The historical task quality score factor reflects the comprehensive index of the data quality of the task points completed by the candidate robot in the past. The current task load factor reflects the number of tasks remaining for the candidate robot or the estimated remaining working time.

4. The low-battery inspection task processing system for an inspection robot according to claim 1 or 3, characterized in that: The vehicle control unit also includes calling a preset dynamic energy consumption prediction model to evaluate the reachability of all task points in the remaining task list one by one. The reachability reflects the total estimated energy consumption required to complete the remaining task points sequentially from the current location. It is compared with a preset available operating energy, which reflects the power difference between the first node and a preset second node below the first node. The return safety margin dynamically calculated based on the distance from the current location to the charging station is deducted. If the evaluation result is that the available operating energy is sufficient to complete at least one task point in the remaining task list, the maximum number of task points that can be completed before the power is exhausted to the second node is determined as the final execution sequence. The remaining, uncompleted task points are packaged into a task package to be handed over and sent to the central collaborative processing server.

5. The low-battery inspection task processing system for an inspection robot according to claim 1, characterized in that: The task exchange protocol includes preset verification conditions, which include: The feasibility of energy consumption swapping was verified, and the current remaining power of the inspection robot that triggered the low power warning signal was greater than the estimated energy consumption required to complete the remaining task package of another inspection robot in a non-low power state. Global time efficiency verification: The sum of the time taken for the inspection robot that triggered the low battery warning signal and another inspection robot that was not in a low battery state to move from their current positions to each other's task starting point is less than or equal to the preset time window threshold. Sensor capability compatibility verification: The sensor configurations of the inspection robot that triggers the low battery warning signal and another inspection robot that is not in a low battery state must be able to meet the data collection requirements of all task points in the other's task package.

6. The low-battery inspection task processing system for an inspection robot according to claim 5, characterized in that: When it is determined that there is an exchange object that meets all the verification conditions, the central collaborative processing server simultaneously issues a task exchange instruction to the inspection robot that triggered the low battery warning signal and the exchange object. The task exchange instruction includes the remaining task package information of the other party and the optimal path planning to reach the task starting point. If there are differences in the sensor model or calibration parameters of the same type between the inspection robot that triggered the low battery warning signal and the exchange object, the task exchange instruction also includes a data acquisition parameter conversion matrix.

7. The low-battery inspection task processing system for an inspection robot according to claim 1, characterized in that: The central collaborative processing server also includes an initial task allocation module. The initial task allocation module uses a density-based spatial clustering algorithm to cluster the three-dimensional coordinates of all task points to be executed, forming several spatially adjacent task point clusters. It calculates the total task amount for each task point cluster, which reflects the sum of the baseline time estimate of all task points in the cluster and the estimated movement time of the shortest traversal path between task points in the cluster. Then, according to the optimization allocation algorithm, the task point clusters are allocated to different inspection robots to minimize the variance of the expected remaining battery power after all robots complete their respective tasks, thereby forming their respective initial task lists.

8. The low-battery inspection task processing system for an inspection robot according to claim 1, characterized in that: The central collaborative processing server also includes a data fusion and report generation module. The data fusion and report generation module accurately correlates all data collected by different inspection robots at different times and through different collaborative processing protocols back to the originally defined key detection points based on the globally unique task point identifier attached to each data file, and generates a comprehensive inspection report.

9. The low-battery inspection task processing system for an inspection robot according to claim 1, characterized in that: The inputs to the dynamic energy consumption prediction model include the inherent physical parameters of the inspection robot, real-time kinematic parameters, real-time load parameters, real-time power consumption of the sensor system, and real-time computing load of the on-board computing unit.

10. The low-battery inspection task processing system for an inspection robot according to claim 9, characterized in that: The dynamic energy consumption prediction module calculates the robot's instantaneous total power consumption in real time by weighted summation of drive power, sensor system power, and computing unit power. The drive power reflects a function of real-time kinematic parameters and real-time load parameters, the sensor system power is determined based on the list of activated sensors and their working status, and the computing unit power reflects the real-time load rate of the on-board computing unit.

Images