Self-adaptive knowledge-driven optimization method and system for intelligent inspection of quadruped robot

Abstract

The invention discloses a self-adaptive knowledge-driven optimization method and system for intelligent inspection of a quadruped robot, and particularly relates to the technical field of driving optimization, and the method comprises the steps: obtaining inspection task data of the quadruped robot, and inputting the inspection task data into a pre-constructed power consumption prediction model to obtain a minimum power consumption value; obtaining a decision training data set of the quadruped robot; training a power consumption replacement decision model for deciding whether to replace the power consumption of the inspection task or not on the basis of the decision training data set when it is detected that the power consumption of the quadruped robot is abnormal; judging whether the power consumption of the quadruped robot is abnormal or not based on the minimum power consumption value predicted by the power consumption prediction model; whether to replace the power consumption of the inspection task is determined according to the real-time data, so that the energy waste is reduced to the maximum extent, and the single inspection endurance of the robot is prolonged; in the long-term large-scale inspection operation, the frequency of frequent charging is reduced, the energy cost is reduced, the inspection continuity is guaranteed, and the overall efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of drive optimization, and more specifically, to an adaptive knowledge-driven optimization method and system for intelligent inspection of quadruped robots. Background Art

[0002] With the development of industrial automation and intelligence, quadruped robots are widely used in inspection tasks in complex environments due to their excellent terrain adaptability and flexible motion performance. However, quadruped robots face many challenges in actual operation, and power consumption optimization is a key issue. The robot needs to run for a long time in a complex environment, and the battery life and energy efficiency directly determine the task completion ability and economic benefits. However, uncertain factors in complex environments, such as terrain complexity, obstacle distribution, and task priority changes, often make the power consumption efficiency of traditional methods low, and even lead to task failure.

[0003] In existing methods, for example, the Chinese patent application with publication number CN118656734A discloses a petrochemical area inspection system based on an inspection robot, including: based on the real-time collection of sensor data between inspection points, analyzing the data between multiple inspection points, integrating node dependency and data mobility indicators, and calculating the interaction intensity of each node through a neural network model. Although the above technical solution optimizes the inspection path configuration and improves the inspection efficiency and safety by dynamically integrating and analyzing sensor data to calculate the interaction intensity and dependency relationship between nodes, through research and application of the technical solution and the existing technology, it is found that the above technical solution and the existing technology have at least the following partial defects:

[0004] The robot fails to dynamically adjust power consumption according to task requirements, resulting in insufficient battery life or resource waste.

[0005] Therefore, the present invention provides an adaptive knowledge-driven optimization method and system for intelligent inspection of quadruped robots. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides an adaptive knowledge-driven optimization method and system for intelligent inspection of quadruped robots to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] In the first aspect, the present invention provides an adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots, including:

[0009] Step 1: Obtain the inspection task data of the quadruped robot, and input the inspection task data into a pre-constructed power consumption prediction model to obtain the minimum power consumption value;

[0010] Step 2: Obtain the decision training data set of the quadruped robot; based on the decision training data set, train a power consumption replacement decision model that decides whether to use the replacement inspection task power consumption when detecting abnormal power consumption of the quadruped robot.

[0011] Step 3: Based on the minimum power consumption value predicted by the power consumption prediction model, determine whether there is abnormal power consumption in the quadruped robot; if there is abnormal power consumption, generate a power consumption abnormal instruction and go to Step 4; if there is no abnormal power consumption, return to Step 1.

[0012] Step 4: Receive the power consumption abnormal instruction, obtain the replacement inspection task power consumption, and based on the replacement inspection task power consumption, operating power consumption, and power consumption replacement decision model, output a decision on whether to use the replacement inspection task power consumption.

[0013] Furthermore, the inspection task data includes a motion power consumption score and a task power consumption score.

[0014] The motion power consumption score is obtained by calculating through combining the environmental impact data where the robot is located, motion speed, acceleration, and motion type for assignment.

[0015] Furthermore, the task power consumption score is obtained as follows:

[0016] Step a11: Obtain the task data of the quadruped robot, where the task data includes the detected task type and task execution duration; the detected task types include equipment appearance inspection tasks, equipment operation status inspection tasks, and equipment internal parameter measurement tasks.

[0017] Step a12: Analyze the task data to obtain the task power consumption score.

[0018] Furthermore, the training method of the power consumption prediction model is:

[0019] Divide the historical inspection task training data into a task training set and a task test set, and construct a regression network model. The historical inspection task training data includes inspection task data and the corresponding minimum power consumption value of the inspection task data.

[0020] Taking the inspection task data in the task training set as the input and the corresponding minimum power consumption value as the output, both are input into the regression network model for training to construct the initial regression network model. The initial regression network model selects a neural network model, and the training process focuses on minimizing the total prediction accuracy. After the preliminary training is completed, the task test set is used to evaluate the initial regression network model. If the total prediction accuracy achieved by the initial regression network model in the task test set is lower than the preset threshold, the initial regression network model is established as the power consumption prediction model; otherwise, if the total prediction accuracy reaches or exceeds the preset threshold, the task training set is re-input, and the initial regression network model is continuously iteratively trained until the total prediction accuracy obtained in the test session meets the preset standard.

[0021] Further, the method for obtaining the minimum power consumption value corresponding to the inspection task data in the historical inspection task training data is as follows:

[0022] Step b1: Obtain the robot mass and the total running duration.

[0023] Step b2: Perform a formula-based calculation on the inspection task data of the m-th industrial device, the robot mass, and the total running duration to obtain the power consumption evaluation score of the m-th industrial device; m = 1, 2,..., M; M is the total number of industrial devices.

[0024] Step b3: Let m = m + 1, repeat the above steps until m = M, then end the loop, and mark the minimum power consumption evaluation score corresponding to the industrial device as the minimum power consumption value.

[0025] Further, the method for obtaining the decision training data set includes:

[0026] Step c1: Obtain the replacement decision data of the quadruped robot. The replacement decision data includes the power consumption value of the m-th industrial device, the task completion index, the power consumption of the replacement inspection task, the robot mass, and the task power consumption replacement decision.

[0027] Step c2: Generate a decision training data set based on the replacement decision data.

[0028] The method for obtaining the task completion index includes:

[0029] Obtain the task completion feature data of the m-th industrial device. The task completion feature data includes the inspection task completion rate and the data transmission completion rate.

[0030] The inspection task completion rate is the ratio of the number of tasks that the quadruped robot has inspected and completed for the m-th industrial device to the preset number of inspected and completed tasks; the data transmission completion rate is the ratio of the data transmission for the quadruped robot to complete the inspection task for the m-th industrial device to the preset data transmission.

[0031] Determine the task completion status according to the task completion index, where the task completion status includes that the inspection task is not completed and the inspection task is completed;

[0032] The method for determining the task completion status according to the task completion index includes:

[0033] Preset a task index threshold, and compare the task completion index with the task index threshold;

[0034] When the task completion index is less than or equal to the task index threshold, generate the task completion status as the inspection task not completed;

[0035] When the task completion index is greater than the task index threshold, generate the task completion status as the inspection task completed;

[0036] The replacement inspection task power consumption is the power consumption of stopping the inspection task and returning to the starting point;

[0037] The task power consumption replacement decision is that each time the quadruped robot is in the inspection task process, when facing the power consumption state, it chooses or does not choose to use one of the replacement inspection task power consumptions.

[0038] Furthermore, the decision training data set refers to generating a set of decision training data corresponding to each group of replacement decision data; generating a decision training data set based on all replacement decision data;

[0039] The decision training data includes the current state, the selected action, the reward value, and the next state corresponding to each group of replacement decision data;

[0040] The method for generating a decision training data set based on replacement decision data is:

[0041] Take the robot's battery level, the power consumption value of each group of replacement decision data, and the replacement inspection task power consumption as the current state;

[0042] Take the task power consumption replacement decision in the replacement decision data as the selected action;

[0043] Calculate the reward value Q after each group of replacement decision data selects the action;

[0044] Take the operating power consumption of the next group of replacement decision data as the next state.

[0045] Furthermore, the method for training a power consumption replacement decision model that decides whether to use the replacement inspection task power consumption when detecting abnormal power consumption of the quadruped robot includes:

[0046] Use the current state and reward value in the decision training data set as the power consumption to replace the input of the decision model. The power consumption replacement decision model randomly extracts multiple groups of decision training data from the decision training data set for training, and learns the strategy of whether to choose to use the power consumption of the replacement inspection task under different power consumption states to obtain the maximum reward value; the power consumption replacement decision model is a deep learning model.

[0047] Further, the method for judging whether there is abnormal power consumption in the quadruped robot includes:

[0048] Obtain the remaining power consumption value of the quadruped robot, calculate the difference between the lowest power consumption value predicted by the power consumption prediction model and the remaining power consumption value to obtain the power consumption difference;

[0049] Compare the power consumption difference with the preset power consumption difference threshold;

[0050] If the power consumption difference is less than the preset power consumption difference threshold, generate a power consumption abnormal instruction;

[0051] If the power consumption difference is greater than or equal to the preset power consumption difference threshold, do not generate a power consumption abnormal instruction;

[0052] Among them, the method for outputting the decision on whether to use the power consumption of the replacement inspection task includes:

[0053] Input the obtained lowest power consumption value, the generated power consumption of the replacement inspection task, and the robot mass of the quadruped robot to be controlled into the power consumption replacement decision model, and output the decision on whether to choose to use the power consumption of the replacement inspection task.

[0054] In the second aspect, the present invention provides an adaptive knowledge-driven optimization system for intelligent inspection of quadruped robots; used to implement the above-mentioned adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots, including:

[0055] A data acquisition module, used to obtain the inspection task data of the quadruped robot, and input the inspection task data into a pre-constructed power consumption prediction model to obtain the lowest power consumption value;

[0056] A decision model training module, used to obtain the decision training data set of the quadruped robot; based on the decision training data set, train a power consumption replacement decision model that decides whether to use the power consumption of the replacement inspection task when detecting abnormal power consumption of the quadruped robot;

[0057] A judgment module, based on the lowest power consumption value predicted by the power consumption prediction model, judges whether there is abnormal power consumption in the quadruped robot; if there is abnormal power consumption, generate a power consumption abnormal instruction and transfer it to the decision module; if there is no abnormal power consumption, return to the data acquisition module;

[0058] A decision-making module, configured to receive a power consumption anomaly instruction, obtain the power consumption of the replacement inspection task, and output a decision on whether to use the power consumption of the replacement inspection task based on the power consumption of the replacement inspection task, the operating power consumption, and the power consumption replacement decision-making model.

[0059] The technical effects and advantages of the present invention:

[0060] 1. Through the accurate power consumption prediction model of the present invention, the lowest power consumption value is calculated in advance, enabling the inspection task planning to be based on evidence and avoiding unnecessary high-power consumption paths. When a power consumption anomaly is detected, the power consumption replacement decision-making model quickly intervenes and decides whether to enable the power consumption of the replacement inspection task based on real-time data, minimizing energy waste to the greatest extent and extending the single-inspection endurance of the robot. In long-term large-scale inspection operations, the number of frequent charging times is reduced, not only reducing energy costs but also ensuring the continuity of inspections and improving overall efficiency.

[0061] 2. The present invention relies on the detailed ways of obtaining the task power consumption score and the motion power consumption score to quantify the power consumption in line with the actual scenario. During the task execution, the progress is dynamically tracked through the task completion index, the completion status of the inspection task is accurately judged, and adjustments are made in a timely manner when a power consumption anomaly occurs to ensure that key inspection links are not missed. The power consumption replacement decision-making model of deep learning continuously learns and optimizes, enabling the robot to make scientific power consumption decisions under complex working conditions, effectively reducing inspection interruptions caused by power consumption problems, and comprehensively enhancing the accuracy and integrity of inspection data. Description of the Drawings

[0062] Figure 1 Flowchart of the adaptive knowledge-driven optimization method for the intelligent inspection of a quadruped robot in Embodiment 1;

[0063] Figure 2 Flowchart of the method for obtaining the lowest power consumption value corresponding to the inspection task data in the historical inspection task training data of Embodiment 1;

[0064] Figure 3 Flowchart of the method for determining whether there is a power consumption anomaly in a quadruped robot in Embodiment 1;

[0065] Figure 4 Structural schematic diagram of the adaptive knowledge-driven optimization system for the intelligent inspection of a quadruped robot in Embodiment 2. Detailed Embodiments

[0066] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities may be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0068] It should be understood that although terms such as "first" and "second" may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0069] Embodiment 1

[0070] Please refer to Figure 1 As shown, the present embodiment discloses an adaptive knowledge-driven optimization method for intelligent inspection of a quadruped robot, which is applied to a quadruped robot. The method includes:

[0071] Step 1: Obtain the inspection task data of the quadruped robot, and input the inspection task data into a pre-constructed power consumption prediction model to obtain the minimum power consumption value;

[0072] It should be noted that: the inspection object of the quadruped robot may be industrial equipment or power equipment, etc. In this embodiment, the inspection task of M industrial equipment by the quadruped robot in the target factory is taken as an example for illustration. Each industrial equipment inspection task includes a motion task and a detection task. The motion task includes walking, jumping, and running; the detection task includes an equipment appearance inspection task, an equipment operation status inspection task, and an equipment internal parameter measurement task.

[0073] The inspection task data includes a motion power consumption score and a task power consumption score.

[0074] Among them, the obtaining method of the motion power consumption score includes:

[0075] Step a01: Obtain the basic motion parameters of the quadruped robot. The basic motion parameters include a motion type assignment, the mass of the robot, the motion speed, and the acceleration;

[0076] It should be noted that: the motion type assignment is preset, and different values are set according to the motion state of the quadruped robot. The motion type assignment corresponding to the quadruped robot in the walking state is set to 1, the motion type assignment corresponding to the quadruped robot in the jumping state is set to 2, and the motion type assignment corresponding to the quadruped robot in the running state is set to 3; the motion speed is the average forward speed of the quadruped robot within a preset time period, which is calculated by the ratio of displacement to the duration of the preset time period; the acceleration is the speed change of the quadruped robot. The motion speed and acceleration are obtained by collecting through a speed sensor.

[0077] Step a02: Perform a formula calculation on the basic motion parameters and the standby power consumption of the quadruped robot to obtain a motion power consumption score, and its calculation formula is:

[0078]

[0079] In the formula, E yd represents the motion power consumption score, ZL represents the robot mass, Hy represents the environmental impact data, BT represents the motion type assignment, SD represents the motion speed, DX represents the acceleration, and P0 represents the standby power consumption of the quadruped robot.

[0080] It should be noted that: the environmental impact data is obtained by the quadruped robot using devices such as lidar and ultrasonic sensors to scan the terrain around the robot. By analyzing parameters such as the undulation degree and roughness of the terrain, the influence degree of the environment on the robot's motion is determined. It is preset by those skilled in the art according to the geographical environment. Exemplarily, if the quadruped robot moves on a flat ground and the terrain sensor detects a small terrain change, the environmental impact data is set to a small value, such as 0.2; while when the quadruped robot is in a rugged mountain environment with large terrain undulations and the terrain parameters detected by the sensor change violently, the environmental impact data increases to 0.8.

[0081] The greater the mass of the robot and the heavier the load, the greater the gravitational potential energy that the robot needs to overcome, and the corresponding increase in motion energy consumption, which is reflected by to reflect its influence degree on energy consumption.

[0082] The acquisition method of the task power consumption score includes:

[0083] Step a11: Obtain the task data of the quadruped robot, and the task data includes the detection task type power consumption and the task execution duration;

[0084] It should be noted that: the detection task type power consumption includes the appearance detection power consumption P1 and the internal parameter measurement power consumption P2.

[0085] Among them, the acquisition of the appearance detection power consumption P1 is achieved by, when the appearance inspection task is executed, the power sensor equipped on the robot records the power data in real time, and the total energy consumption P1 is calculated through time integration. Its calculation formula is: t1 and t2 represent the start and end times of the appearance inspection task, represents the real-time power measured by the energy consumption sensor during the appearance inspection process. The monitoring method of the internal parameter measurement power consumption P2 is the same as that of the appearance detection power consumption P1, and will not be repeated here.

[0086] Step a12: Analyze the task data to obtain the task power consumption score. Its fitting formula is:

[0087]

[0088] In the formula, E rw represents the task power consumption score, T r represents the task execution time, and φ1 and φ2 are the corresponding weight data, and φ1 + φ2 = 1.

[0089] In implementation, the training method of the power consumption prediction model is:

[0090] Divide the historical inspection task training data into a task training set and a task test set, and construct a regression network model. The historical inspection task training data includes inspection task data and the corresponding minimum power consumption value of the inspection task data; the inspection task training data is obtained by analyzing the tested quadruped robot in different inspection tasks.

[0091] Please refer to Figure 2 As shown, among them, the method for obtaining the minimum power consumption value corresponding to the inspection task data in the historical inspection task training data is:

[0092] Step b1: Obtain the robot mass and the total running duration;

[0093] It should be noted that: the robot mass is previously measured by a gravity sensor; the total running duration is through the timer built in the device. Since the battery starts to supply power to the quadruped robot, the cumulative timing function is started synchronously until the battery runs out of power and the quadruped robot stops running. The final cumulative data of the timer is the usage duration of the battery during this complete power supply process.

[0094] Step b2: Perform formula calculation on the inspection task data, robot mass, and total running duration of the mth industrial device to obtain the power consumption evaluation score of the mth industrial device; m = 1, 2,..., M; its calculation formula is:

[0095] EP m =(E yd +E rw)×α1 + ZL×α2 + DCZ×α3;

[0096] In the formula, EP m represents the power consumption evaluation score of the m-th industrial device, DCZ represents the total running duration, and α1, α2, and α3 are all weight data, with α1 + α2 + α3 = 1.

[0097] Step b3: Let m = m + 1, repeat the above steps until m = M, then end the loop, and mark the minimum power consumption evaluation score corresponding to the industrial device as the minimum power consumption value.

[0098] Taking the inspection task data in the task training set as the input and the corresponding minimum power consumption value as the output, inputting both into the regression network model for training to construct an initial regression network model. The initial regression network model selects a neural network model, and the training process focuses on minimizing the total prediction accuracy. After completing the preliminary training, use the task test set to evaluate the initial regression network model. If the total prediction accuracy achieved by the initial regression network model in the task test set is lower than the preset threshold, then establish the initial regression network model as the power consumption prediction model; otherwise, if the total prediction accuracy reaches or exceeds the preset threshold, then re-input the task training set and continue to iteratively train the initial regression network model until the total prediction accuracy obtained in the test session meets the preset standard.

[0099] In this step, the quadruped robot predicts the power consumption during the inspection of M industrial devices through the inspection task data, for the next step of adjusting the order of the inspection tasks. Place the simple appearance inspection tasks with low power consumption at a slightly lower battery level, and arrange high-power consumption tasks such as in-depth scanning of complex equipment and multi-sensor collaborative detection when the battery is full, making the execution rhythm of the quadruped robot's inspection tasks more reasonable.

[0100] Step 2: Obtain the decision training data set of the quadruped robot; train a power consumption replacement decision model based on the decision training data set to decide whether to use the power consumption of the replacement inspection task when detecting abnormal power consumption of the quadruped robot;

[0101] In implementation, the method for obtaining the decision training data set includes:

[0102] Step c1: Obtain the replacement decision data of the quadruped robot, where the replacement decision data includes the power consumption value of the m-th industrial device, the task completion index, the power consumption of the replacement inspection task, the robot mass, and the task power consumption replacement decision;

[0103] Among them, the method for obtaining the task completion index includes:

[0104] Obtain the task completion feature data of the m-th industrial device, where the task completion feature data includes the inspection task completion rate and the data transmission completion rate;

[0105] It should be noted that: the inspection task completion rate is the ratio between the number of tasks that the quadruped robot has completed the inspection of the m-th industrial device and the preset number of completed inspection tasks; the data transmission completion rate is the ratio between the data transmission of the quadruped robot completing the inspection task of the m-th industrial device and the preset data transmission.

[0106]

[0107] In the formula, RW represents the task completion index, P sx represents the inspection equipment completion rate, P sc represents the data transmission completion rate.

[0108] It should be noted that: in the numerator part, when P sx ×P sc tends to 1, the value of the logarithmic function rises rapidly, more sensitively reflecting the advantages of a high completion rate; when P sx ×P sc is close to 0, the growth is slow; in the denominator part, when P sx ×P sc is closer to 1, the denominator tends to 0, and the value of the whole fraction is larger, that is, the task completion index is higher; on the contrary, when the completion rates of both items are very low, the denominator is close to 1, and the improvement effect on the overall value is limited, thus highlighting the disadvantages when the task completion situation is poor.

[0109] Determine the task completion status according to the task completion index, where the task completion status includes the inspection task not completed and the inspection task completed;

[0110] In implementation, the method for determining the task completion status according to the task completion index includes:

[0111] Preset a task index threshold, compare the task completion index with the task index threshold. When the task completion index is less than or equal to the task index threshold, it indicates that the inspection of the m-th industrial device is not completed, and the task completion status is generated as the inspection task not completed; when the task completion index is greater than the task index threshold, it indicates that the inspection of the m-th industrial device is completed, and the task completion status is generated as the inspection task completed;

[0112] The larger the task completion index is, the higher the inspection task completion rate of the corresponding m-th industrial equipment inspection is. Then the quadruped robot does not use the power consumption of the replacement inspection task. This is because the quadruped robot has been deeply optimized and adapted to the existing inspection task process. Even if the power consumption of the new replacement inspection task is low, there may be compatibility problems with the existing system in links such as instruction issuance, action execution, and data transmission, resulting in unstable operation and increased error probability of the quadruped robot. Therefore, for reliability considerations, the original high-power and high-completion-rate task is still used.

[0113] The power consumption of the replacement inspection task is the power consumption of stopping the inspection task and returning to the starting point.

[0114] The method for obtaining the power consumption of the replacement inspection task includes:

[0115] Step d1: Obtain the path to return to the starting point:

[0116]

[0117] In the formula, d represents the path to return to the starting point, x c and y c represent the current coordinate point of the quadruped robot, x s and y s represent the starting point coordinate of the quadruped robot;

[0118] Step d2: Obtain the operation switching power consumption:

[0119] E qh = P qh × T qh ;

[0120] In the formula, E qh represents the switching power consumption, P qh represents the operation switching power, T qh represents the operation switching time.

[0121] It should be noted that: the operation switching power is obtained through a power sensor.

[0122] Step d3: Obtain the return speed, and comprehensively analyze the return speed, the path to return to the starting point, the motion power consumption score, and the switching power consumption to obtain the power consumption of the replacement inspection task. The calculation formula is:

[0123]

[0124] In the formula, E gh represents the power consumption of the replacement inspection task, V fh represents the return speed.

[0125] The task power consumption replacement decision means that every time the quadruped robot is in the inspection task process, when the quadruped robot faces the power consumption state, it chooses or does not choose to use one of the replacement inspection task power consumptions; if the power consumption is abnormal, the replacement inspection task power consumption is replaced, and if the power consumption is normal, the replacement inspection task power consumption is not replaced.

[0126] Step c2: Generate a decision training data set based on the replacement decision data.

[0127] It should be noted that: the decision training data set refers to generating a set of decision training data corresponding to each group of replacement decision data; generating a decision training data set based on all replacement decision data;

[0128] The decision training data includes the current state, the selected action, the return value, and the next state corresponding to each group of replacement decision data;

[0129] In implementation, the acquisition method of generating a decision training data set based on the replacement decision data is:

[0130] Taking the robot's battery power, the power consumption value of each group of replacement decision data, and the replacement inspection task power consumption as the current state;

[0131] Taking the task power consumption replacement decision in the replacement decision data as the selected action;

[0132] Calculate the return value after each group of replacement decision data selects the action; among them, the calculation method of the return value is:

[0133] Mark the power consumption value of the mth industrial device as K, mark the robot mass as ZL, and mark the replacement inspection task power consumption as XS; then the calculation formula of the return value Q is:

[0134] Q = h × [γ × (K - XS) - θ × ZL + μ × XS + RW] + C × (1 - h);

[0135] Among them, h = 0 or h = 1. Exemplarily, when the task power consumption replacement decision is not to choose to use the replacement inspection task power consumption, then h = 0; when the task power consumption replacement decision is to choose to use the replacement inspection task power consumption, then h = 1. In the formula, γ represents the preset proportional score, θ represents the weight data of the robot mass, μ represents the weight data of the replacement inspection task power consumption, and C is a constant correction score, which is corrected by the staff according to experience.

[0136] It can be understood that when K - XS is smaller, that is, when the difference between the power consumption value and the replacement inspection task power consumption is smaller, when the quadruped robot has normal power consumption, it is more inclined to h = 0, that is, more inclined not to choose to replace the inspection task power consumption; when RW is smaller, it means that the task completion index is smaller, then when the quadruped robot has abnormal power consumption, it makes h = 1 more, that is, chooses to use the replacement inspection task power consumption.

[0137] The operating power consumption of the following set of replacement decision data is for the next state.

[0138] In implementation, the method for training a power consumption replacement decision model that decides whether to use the power consumption of the replacement inspection task when detecting abnormal power consumption of a quadruped robot includes:

[0139] Taking the current state and the reward value in the decision training data set as the input of the power consumption replacement decision model, the power consumption replacement decision model randomly extracts multiple groups of decision training data from the decision training data set for training, and learns the strategy of whether to choose to use the power consumption of the replacement inspection task under different power consumption states to obtain the maximum reward value; the power consumption replacement decision model is a deep learning model.

[0140] Step 3: Based on the lowest power consumption value predicted by the power consumption prediction model, determine whether there is abnormal power consumption in the quadruped robot; if there is abnormal power consumption, generate a power consumption abnormal instruction and go to Step 4; if there is no abnormal power consumption, return to Step 1.

[0141] Please refer to Figure 3 As shown, in implementation, the method for determining whether there is abnormal power consumption in a quadruped robot includes:

[0142] Obtain the remaining power consumption value of the quadruped robot, calculate the difference between the lowest power consumption value predicted by the power consumption prediction model and the remaining power consumption value to obtain a power consumption difference;

[0143] It should be noted that the remaining power consumption value includes the sum of the safety remaining amount and the remaining power consumption amount; the safety remaining amount is the lowest power consumption required for the quadruped robot to safely return to the initial position or the charging station.

[0144] Compare the power consumption difference with a preset power consumption difference threshold;

[0145] If the power consumption difference is less than the preset power consumption difference threshold, generate a power consumption abnormal instruction, indicating that the quadruped robot cannot complete the inspection task of the mth industrial device;

[0146] If the power consumption difference is greater than or equal to the preset power consumption difference threshold, do not generate a power consumption abnormal instruction, indicating that the quadruped robot can complete the inspection task of the mth industrial device.

[0147] Step 4: Receive the power consumption abnormal instruction, obtain the power consumption of the replacement inspection task, and based on the power consumption of the replacement inspection task, the operating power consumption, and the power consumption replacement decision model, output a decision on whether to use the power consumption of the replacement inspection task;

[0148] In implementation, the method for outputting a decision on whether to use the power consumption of the replacement inspection task includes:

[0149] Input the obtained minimum power consumption value, the generated power consumption of the replacement inspection task, and the power consumption of the robot mass to be controlled into the power consumption replacement decision model to obtain the decision on whether to select the power consumption of the replacement inspection task as the output.

[0150] In this embodiment, through an accurate power consumption prediction model, the minimum power consumption value is calculated in advance, making the inspection task planning well-founded and avoiding unnecessary high-power consumption paths. When power consumption anomalies are detected, the power consumption replacement decision model intervenes quickly and decides whether to enable the power consumption of the replacement inspection task based on real-time data, minimizing energy waste to the greatest extent and extending the single inspection endurance of the robot. In long-term large-scale inspection operations, the frequency of charging is reduced, not only reducing energy costs but also ensuring the continuity of inspections and improving overall efficiency.

[0151] This embodiment relies on detailed ways of obtaining task power consumption scores and motion power consumption scores to quantify power consumption in line with the actual scenario. During task execution, the progress is dynamically tracked through the task completion index, the completion status of the inspection task is accurately judged, and adjustments are made in a timely manner when power consumption anomalies are encountered to ensure that key inspection links are not missed. The power consumption replacement decision model based on deep learning continuously learns and optimizes, enabling the robot to make scientific power consumption decisions under complex working conditions, effectively reducing inspection interruptions caused by power consumption problems, and comprehensively enhancing the accuracy and integrity of inspection data.

[0152] Embodiment 2

[0153] Please refer to Figure 4 As shown, this embodiment provides an adaptive knowledge-driven optimization system for intelligent inspection of quadruped robots. The system includes: a data acquisition module, a decision model training module, a judgment module, and a decision module; each module is connected by wired and / or wireless means to achieve data transmission between modules;

[0154] The data acquisition module is used to obtain the inspection task data of the quadruped robot and input the inspection task data into a pre-constructed power consumption prediction model to obtain the minimum power consumption value;

[0155] The decision model training module is used to obtain the decision training data set of the quadruped robot; based on the decision training data set, a power consumption replacement decision model is trained to decide whether to use the power consumption of the replacement inspection task when the power consumption of the quadruped robot is detected to be abnormal;

[0156] The judgment module judges whether there is a power consumption anomaly in the quadruped robot based on the minimum power consumption value predicted by the power consumption prediction model; if there is a power consumption anomaly, a power consumption anomaly instruction is generated and transferred to the decision module; if there is no power consumption anomaly, it returns to the data acquisition module;

[0157] A decision-making module, configured to receive a power consumption anomaly instruction, obtain the power consumption of the replacement inspection task, and output a decision on whether to use the power consumption of the replacement inspection task based on the power consumption of the replacement inspection task, the operating power consumption, and a power consumption replacement decision-making model.

[0158] For the formulas involved above, after removing the dimension, only numerical values are used for calculation. These formulas are obtained by collecting a large amount of data and with the aid of software simulation, and are the formulas closest to the actual situation. The weight data in the formulas, as well as various preset thresholds in the analysis process, are set by professionals in the field according to the actual situation or derived through a large amount of data simulation.

[0159] The main function of the weight data is to quantify each parameter, so as to obtain specific numerical values for subsequent comparison and analysis. The value of the weight data depends on the scale of the sample data and the processing scores preset by technicians for each group of sample data. When setting the weight data, as long as it is ensured that it will not affect the proportional relationship between the parameter and the quantified numerical value.

[0160] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

[0161] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots, characterized in that, Including: Step 1: Obtain the inspection task data of the quadruped robot, and input the inspection task data into a pre-constructed power consumption prediction model to obtain the minimum power consumption value; Step 2: Obtain the decision training data set of the quadruped robot; based on the decision training data set, train a power consumption replacement decision model that decides whether to use the replacement inspection task power consumption when detecting abnormal power consumption of the quadruped robot; Step 3: Based on the minimum power consumption value predicted by the power consumption prediction model, determine whether there is abnormal power consumption in the quadruped robot; if there is abnormal power consumption, generate a power consumption abnormal instruction and go to Step 4; if there is no abnormal power consumption, return to Step 1; Step 4: Receive the power consumption abnormal instruction, obtain the replacement inspection task power consumption, and based on the replacement inspection task power consumption, operating power consumption, and power consumption replacement decision model, output a decision on whether to use the replacement inspection task power consumption.

2. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 1, characterized in that The inspection task data includes a motion power consumption score and a task power consumption score; The acquisition method of the motion power consumption score includes: calculating by combining the environmental impact data, motion speed, acceleration, and motion type in which the robot is located for assignment.

3. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 2, characterized in that, The acquisition method of the task power consumption score includes: Step a11: Obtain the task data of the quadruped robot, and the task data includes the detection task type and the task execution duration; the detection task type includes the device appearance inspection task, the device operation status inspection task, and the device internal parameter measurement task; Step a12: Analyze the task data to obtain the task power consumption score.

4. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 3, characterized in that The training method of the power consumption prediction model is: Divide the historical inspection task training data into a task training set and a task test set, and construct a regression network model. The historical inspection task training data includes the inspection task data and the corresponding minimum power consumption value of the inspection task data; Using the inspection task data in the task training set as the input and the corresponding minimum power consumption value as the output, input the two into the regression network model for training to construct an initial regression network model. The initial regression network model selects a neural network model, and the training process focuses on minimizing the total prediction accuracy. After the preliminary training is completed, use the task test set to evaluate the initial regression network model. If the total prediction accuracy achieved by the initial regression network model in the task test set is lower than the preset threshold, then establish the initial regression network model as the power consumption prediction model; otherwise, if the total prediction accuracy reaches or exceeds the preset threshold, re-input the task training set and continuously iterate and train the initial regression network model until the total prediction accuracy obtained in the test link meets the preset standard.

5. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 4, wherein The method for obtaining the minimum power consumption value corresponding to the inspection task data in the historical inspection task training data is: Step b1: Obtain the robot mass and the total running duration; Step b2: Perform a formula calculation on the inspection task data of the m-th industrial device, the robot mass, and the total running duration to obtain the power consumption evaluation score of the m-th industrial device; m = 1, 2,..., M; M is the total number of industrial devices; Step b3: Let m = m + 1, repeat the above steps until m = M, end the loop, and mark the minimum power consumption evaluation score corresponding to the industrial device as the minimum power consumption value.

6. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 5, characterized in that The method for obtaining the decision training data set includes: Step c1: Obtain the replacement decision data of the quadruped robot, where the replacement decision data includes the power consumption value of the m-th industrial device, the task completion index, the power consumption of the replacement inspection task, the robot mass, and the task power consumption replacement decision; Step c2: Generate a decision training data set based on the replacement decision data; The method for obtaining the task completion index includes: Obtain the task completion feature data of the m-th industrial device, where the task completion feature data includes the inspection task completion rate and the data transmission completion rate; The inspection task completion rate is the ratio between the number of inspection tasks completed by the quadruped robot for the m-th industrial device and the preset number of inspection tasks to be completed; the data transmission completion rate is the ratio between the data transmission of the inspection task completed by the quadruped robot for the m-th industrial device and the preset data transmission; Determine the task completion status according to the task completion index, and the task completion status includes the inspection task not completed and the inspection task completed; The method for determining the task completion status according to the task completion index includes: Preset a task index threshold, and compare the task completion index with the task index threshold; When the task completion index is less than or equal to the task index threshold, generate the task completion status as the inspection task not completed; When the task completion index is greater than the task index threshold, generate the task completion status as the inspection task completed; The power consumption of the replacement inspection task is the power consumption of stopping the inspection task and returning to the starting point; The task power consumption replacement decision is that each time the quadruped robot is in the inspection task process, when the quadruped robot faces the power consumption state, it chooses or does not choose to use one of the power consumptions of the replacement inspection task.

7. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 6, characterized in that, The decision training data set refers to a set of decision training data corresponding to each set of replacement decision data; generate a decision training data set based on all replacement decision data; The decision training data includes the current state, the selected action, the reward value, and the next state corresponding to each set of replacement decision data; The method for generating a decision training data set based on the replacement decision data is: Use the robot battery power, the power consumption value of each set of replacement decision data, and the power consumption of the replacement inspection task as the current state; Use the task power consumption replacement decision in the replacement decision data as the selected action; Calculate the reward value Q after each set of replacement decision data selects the action; Use the operating power consumption of the next set of replacement decision data as the next state.

8. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 7, characterized in that The method for training a power consumption replacement decision model that decides whether to use the power consumption of the replacement inspection task when detecting abnormal power consumption of the quadruped robot includes: Use the current state and the reward value in the decision training data set as the input of the power consumption replacement decision model. The power consumption replacement decision model randomly extracts multiple sets of decision training data from the decision training data set for training, and learns whether to choose to use the power consumption of the replacement inspection task under different power consumption states to obtain the strategy with the maximum reward value; the power consumption replacement decision model is a deep learning model.

9. The adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots according to claim 8, characterized in that, The method for determining whether the quadruped robot has abnormal power consumption includes: Obtain the remaining power consumption value of the quadruped robot, calculate the difference between the minimum power consumption value predicted by the power consumption prediction model and the remaining power consumption value to obtain the power consumption difference; Compare the power consumption difference with a preset power consumption difference threshold; If the power consumption difference is less than the preset power consumption difference threshold, generate a power consumption anomaly instruction; If the power consumption difference is greater than or equal to the preset power consumption difference threshold, do not generate a power consumption anomaly instruction; Among them, the method for outputting a decision on whether to use the power consumption of the replacement inspection task includes: Input the power consumption of the replacement inspection task, the minimum power consumption value, and the robot mass input power consumption into the replacement decision model to output a decision on whether to select the power consumption of the replacement inspection task.

10. An adaptive knowledge-driven optimization system for intelligent inspection of quadruped robots, which is used to implement the adaptive knowledge-driven optimization method for intelligent inspection of quadruped robots described in any one of claims 1-9, characterized in that, Include: A data acquisition module for obtaining the inspection task data of the quadruped robot and inputting the inspection task data into a pre-built power consumption prediction model to obtain the minimum power consumption value; A decision model training module for obtaining a decision training data set of the quadruped robot; training a power consumption replacement decision model that decides whether to use the power consumption of the replacement inspection task when detecting a power consumption anomaly of the quadruped robot based on the decision training data set; A judgment module for judging whether there is a power consumption anomaly of the quadruped robot based on the minimum power consumption value predicted by the power consumption prediction model; If there is a power consumption anomaly, generate a power consumption anomaly instruction and transfer it to the decision module; if there is no power consumption anomaly, return to the data acquisition module; A decision module for receiving the power consumption anomaly instruction, obtaining the power consumption of the replacement inspection task, and outputting a decision on whether to use the power consumption of the replacement inspection task based on the power consumption of the replacement inspection task, the operating power consumption, and the power consumption replacement decision model.

Images