Multi-legged Robot Motion Planning Method Based on Brain-inspired Decision-making and Visual Localization

Abstract

The present invention discloses a motion planning method for a multi-legged robot based on brain-inspired decision-making and visual positioning, belonging to the technical field of multi-legged robots, including: collecting and processing real-time data of the multi-legged robot's motion; constructing a motion planning decision-making model for the multi-legged robot to analyze the motion characteristic data of the multi-legged robot and determine the motion planning decision-making scheme; planning the motion of the multi-legged robot, controlling the gait and actions of the multi-legged robot, and forming a closed-loop control for the motion planning of the multi-legged robot through real-time monitoring. The present invention solves the problem that the existing technology cannot effectively plan the motion of the multi-legged robot based on brain-inspired decision-making and visual positioning, resulting in poor autonomous navigation and motion planning decision-making capabilities of the multi-legged robot in complex environments. The present invention can effectively plan the motion of the multi-legged robot based on brain-inspired decision-making and visual positioning, improve the autonomous navigation and motion planning decision-making capabilities of the multi-legged robot in complex environments, and make the motion effect of the multi-legged robot good.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-legged robots, and in particular to a multi-legged robot motion planning method based on brain-like decision-making and visual positioning. Background Art

[0002] A multi-legged robot is a bionic robot that uses a mechanical arm to move on the ground. Since the end of the mechanical arm is generally equipped with an anti-slip device and a contact sensor for contacting the ground, and there is no mechanism for picking up functions, such a mechanical arm is generally called a foot. Common forms include two types of robots: four-legged and six-legged. Among them, the four-legged robot is similar in size and overall shape to a dog, and is commonly known as a robot dog. Compared with other types of robots, the main application advantage of multi-legged robots is that they have better obstacle-crossing capabilities, such as being able to go up and down stairs and climb over more complex obstacles.

[0003] The Chinese patent with publication number CN118583172A discloses a positioning method and system for a multi-legged robot based on an indoor scene, which determines the friction coefficient of the ground according to multiple ground parameters; defines the driving type according to the friction coefficient and the tire coefficient of the multi-legged robot, and the driving type is adapted to the type of ground on which the multi-legged robot drives; defines the driving posture of the multi-legged robot according to the posture type of each driving foot, thereby being compatible with the consideration of the posture types of multiple driving feet, and determines the driving posture of the multi-legged robot under the comprehensive definition of multiple driving feet, so as to trigger the emergency measures of the multi-legged robot according to the driving posture of the multi-legged robot, so as to ensure the anti-fall form of the multi-legged robot and the driving stability of the multi-legged robot, and at the same time, obtains the surrounding environment image collected by the multi-legged robot, and defines multiple environmental features according to the surrounding environment image, and defines the spatial relationship between the multi-legged robot and the multiple environmental features, so as to determine the current position of the multi-legged robot in the indoor scene. However, the patent has the following defects:

[0004] Existing technologies cannot effectively plan the movement of multi-legged robots based on brain-like decision-making and visual positioning, resulting in poor autonomous navigation and motion planning decision-making capabilities of multi-legged robots in complex environments, leading to poor movement effects of multi-legged robots. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-legged robot motion planning method based on brain-like decision-making and visual positioning, which can effectively plan the motion of the multi-legged robot based on brain-like decision-making and visual positioning, improve the autonomous navigation and motion planning decision-making capabilities of the multi-legged robot in complex environments, make the multi-legged robot move well, and solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A motion planning method for a multi-legged robot based on brain-inspired decision-making and visual positioning, including:

[0008] Collecting real-time data on the motion of the multi-legged robot and processing it to determine the motion characteristic data of the multi-legged robot;

[0009] Constructing a motion planning decision model for the multi-legged robot to analyze the motion characteristic data of the multi-legged robot, making a decision on the optimal motion planning of the multi-legged robot, and determining the motion planning decision scheme of the multi-legged robot;

[0010] Planning the motion of the multi-legged robot, controlling the gait and actions of the multi-legged robot, and monitoring in real time to form a closed-loop control for the motion planning of the multi-legged robot.

[0011] Preferably, collecting real-time data on the motion of the multi-legged robot includes:

[0012] Based on a vision sensor, performing real-time monitoring and collection of the surrounding environment of the multi-legged robot to determine the environmental vision information of the multi-legged robot;

[0013] Based on the environmental vision data of the multi-legged robot and combining simultaneous localization and mapping technology, performing visual positioning on the multi-legged robot and constructing an environmental map to determine the positioning location information of the multi-legged robot;

[0014] Based on a lidar, performing real-time monitoring and collection of obstacles in the surrounding environment of the multi-legged robot to determine the distance information of nearby objects of the multi-legged robot;

[0015] Based on an inertial measurement unit, performing real-time monitoring and collection of the acceleration, angular velocity, and attitude angle of the multi-legged robot to obtain the attitude motion information of the multi-legged robot;

[0016] Among them, based on the environmental vision information, positioning location information, distance information of nearby objects, and attitude motion information of the multi-legged robot, the real-time motion data of the multi-legged robot is determined.

[0017] Preferably, processing the real-time motion data of the multi-legged robot includes:

[0018] Cleaning the real-time motion data of the multi-legged robot to remove the noise data that is useless for the motion planning of the multi-legged robot in the real-time motion data of the multi-legged robot;

[0019] Checking the real-time motion data of the multi-legged robot to identify duplicate values, missing values, and outliers in the real-time motion data of the multi-legged robot;

[0020] Remove the duplicate values, missing values, and outliers in the real-time motion data of the multi-legged robot that are useless for the motion planning of the multi-legged robot, fill in the missing values in the real-time motion data of the multi-legged robot that are useful for the motion planning of the multi-legged robot, and replace the outliers that are useful for the motion planning of the multi-legged robot.

[0021] Preferably, processing the real-time motion data of the multi-legged robot further includes:

[0022] Normalize the real-time motion data of the multi-legged robot to convert the real-time motion data of the multi-legged robot into a unified data format, remove the dimensional differences in the real-time motion data of the multi-legged robot, and determine the standardized real-time motion data of the multi-legged robot;

[0023] Extract features from the real-time motion data of the multi-legged robot, extract the features useful for the motion planning of the multi-legged robot from the real-time motion data of the multi-legged robot, and determine the motion feature data of the multi-legged robot.

[0024] Preferably, constructing a motion planning decision model for the multi-legged robot includes:

[0025] According to the motion planning requirements of the multi-legged robot, collect the historical motion data of the multi-legged robot, and divide the collected historical motion data of the multi-legged robot to determine the training set and the test set;

[0026] Based on deep learning technology, use the training set to train the deep learning model, enable the deep learning model to simulate the cognitive mechanism of the human brain and autonomously learn and adapt to complex environments, and optimize the motion planning decision scheme to determine the motion planning decision model of the multi-legged robot based on brain-like decision-making;

[0027] Use the test set to test the motion planning decision model of the multi-legged robot based on brain-like decision-making, and evaluate the decision-making performance of the motion planning decision model of the multi-legged robot based on brain-like decision-making to determine the best motion planning decision model of the multi-legged robot.

[0028] Preferably, determining the best motion planning decision model of the multi-legged robot includes:

[0029] Evaluate whether the motion planning decision model of the multi-legged robot based on brain-like decision-making can achieve the effect of the motion planning decision of the multi-legged robot based on accuracy, recall rate, and F1 score;

[0030] When the motion planning decision-making model of a multi-legged robot based on brain-inspired decision-making fails to achieve the effect of multi-legged robot motion planning decision-making, the parameters of the motion planning decision-making model of the multi-legged robot based on brain-inspired decision-making are adjusted, and the motion planning decision-making model of the multi-legged robot based on brain-inspired decision-making is continuously iteratively optimized until the motion planning decision-making model of the multi-legged robot based on brain-inspired decision-making can achieve the effect of multi-legged robot motion planning decision-making, and the optimal motion planning decision-making model of the multi-legged robot is determined.

[0031] Preferably, analyze the motion characteristic data of the multi-legged robot to determine the motion planning decision-making scheme of the multi-legged robot, including:

[0032] Obtain the optimal motion planning decision-making model of the multi-legged robot and deploy the optimal motion planning decision-making model of the multi-legged robot in the actual motion planning environment of the multi-legged robot;

[0033] Input the motion characteristic data of the multi-legged robot into the optimal motion planning decision-making model of the multi-legged robot, analyze the motion characteristic data of the multi-legged robot according to the optimal motion planning decision-making model of the multi-legged robot, make a decision on the optimal motion planning of the multi-legged robot, and determine the motion planning decision-making scheme of the multi-legged robot;

[0034] Plan the motion of the multi-legged robot according to the motion planning decision-making scheme of the multi-legged robot, control the gait and actions of the multi-legged robot, plan the motion path of the multi-legged robot, so that the multi-legged robot walks smoothly and can avoid obstacles and reach the target position in a complex environment.

[0035] Preferably, determine whether there is an abnormal walking state for the gait parameters during the walking process of the multi-legged robot, and when there is an abnormal walking state, adjust the motion path, including:

[0036] Real-time collect the gait parameters during the walking process of the multi-legged robot, where the gait parameters include the mechanical vibration frequency and the center of gravity deviation value during the robot's walking process;

[0037] Compare the mechanical vibration frequency with a preset mechanical vibration frequency reference value;

[0038] When the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value, the center of gravity position at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value is retrieved;

[0039] Obtain the center of gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value according to the difference degree between the center of gravity position at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the theoretical center of gravity position;

[0040] Utilize the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value;

[0041] Determine whether the current multi-legged robot has an abnormal walking state based on the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value;

[0042] When the walking state is abnormal, adjust the motion path.

[0043] Preferably, determining whether the current multi-legged robot has an abnormal walking state based on the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value includes:

[0044] Retrieve the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value;

[0045] Compare the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value with the preset center-of-gravity deviation value reference value;

[0046] Select the center-of-gravity deviation values that are not lower than the preset center-of-gravity deviation value reference value;

[0047] Retrieve the mechanical vibration frequency of the multi-legged robot corresponding to the moment when the center-of-gravity deviation value is not lower than the preset center-of-gravity deviation value reference value;

[0048] Obtain the standard deviation of the mechanical vibration frequency based on the mechanical vibration frequency of the multi-legged robot corresponding to the moment when the center-of-gravity deviation value is not lower than the preset center-of-gravity deviation value reference value;

[0049] Obtain an abnormal evaluation coefficient by combining the center-of-gravity deviation value that is not lower than the preset center-of-gravity deviation value reference value and the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value with the standard deviation of the mechanical vibration frequency;

[0050] Among them, the abnormal evaluation coefficient is obtained through the following formula:

[0051]

[0052] Among them, Y represents the abnormal evaluation coefficient; Y0 represents the preset coefficient reference value; n represents the number of center-of-gravity deviation values that are not lower than the preset center-of-gravity deviation value reference value; Z i represents the value of the i-th center-of-gravity deviation value that is not lower than the preset center-of-gravity deviation value reference value; Z cRepresents a preset reference value of the center of gravity deviation value; Z represents the center of gravity deviation value at the moment when the mechanical vibration frequency of the robot exceeds the preset reference value of the mechanical vibration frequency; f represents the mechanical vibration frequency of the robot exceeding the preset reference value of the mechanical vibration frequency; f b Represents the standard deviation of the mechanical vibration frequency;

[0053] Compare the abnormal evaluation coefficient with a preset coefficient threshold;

[0054] When the abnormal evaluation coefficient exceeds the preset coefficient threshold, it is determined that there is an abnormality in the walking state of the multi-legged robot.

[0055] Preferably, real-time monitoring forms a closed-loop control for the motion planning of the multi-legged robot, including:

[0056] Real-time monitor the motion of the multi-legged robot, and feed the monitoring data back to the motion planning decision-making model of the multi-legged robot in real time, form a closed-loop control for the motion planning of the multi-legged robot, and then dynamically adjust and optimize the motion planning decision-making scheme of the multi-legged robot, and conduct closed-loop management on the motion planning of the multi-legged robot.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] The present invention collects the environmental vision information, positioning position information, short-distance object distance information and attitude motion information of the multi-legged robot, determines the real-time motion data of the multi-legged robot, determines the motion characteristic data of the multi-legged robot after processing the real-time motion data of the multi-legged robot, constructs a motion planning decision-making model for the multi-legged robot according to the motion planning requirements of the multi-legged robot, analyzes the motion characteristic data of the multi-legged robot, makes a decision on the optimal motion planning of the multi-legged robot, determines the motion planning decision-making scheme of the multi-legged robot, plans the motion of the multi-legged robot according to the motion planning decision-making scheme of the multi-legged robot, controls the gait and actions of the multi-legged robot, and real-time monitors the motion of the multi-legged robot to form a closed-loop control for the motion planning of the multi-legged robot, and can effectively plan the motion of the multi-legged robot based on brain-like decision-making and visual positioning, improve the autonomous navigation and motion planning decision-making ability of the multi-legged robot in a complex environment, and make the motion effect of the multi-legged robot good. Description of the Drawings

[0059] Figure 1 Is a flowchart of the motion planning method for a multi-legged robot based on brain-like decision-making and visual positioning of the present invention. Detailed Embodiments

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to 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 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.

[0061] To solve the problem that the existing technology cannot effectively plan the movement of a multi-legged robot based on brain-inspired decision-making and visual positioning, resulting in poor autonomous navigation and motion planning decision-making capabilities of the multi-legged robot in a complex environment and poor movement effect of the multi-legged robot, please refer to Figure 1 This embodiment provides the following technical solutions:

[0062] A motion planning method for a multi-legged robot based on brain-inspired decision-making and visual positioning, including:

[0063] Collecting real-time data of the multi-legged robot's movement and processing it to determine the movement characteristic data of the multi-legged robot;

[0064] In this embodiment, collecting real-time data of the multi-legged robot's movement includes:

[0065] Based on a visual sensor, the surrounding environment of the multi-legged robot is monitored and collected in real time to determine the environmental visual information of the multi-legged robot for target recognition, obstacle detection, and scene understanding;

[0066] Based on the environmental visual data of the multi-legged robot and combined with the simultaneous localization and mapping technology, the multi-legged robot is visually located and an environmental map is constructed to determine the positioning location information of the multi-legged robot;

[0067] It should be noted that visual positioning is a key link in the motion planning of a multi-legged robot, especially in a complex or dynamic environment. The visual positioning technology obtains environmental information through a camera or other visual sensors and combines the simultaneous localization and mapping technology to help the multi-legged robot determine its own position and construct an environmental map. Through visual perception, the multi-legged robot can identify target objects in the environment, such as obstacles and path signs, and perform tracking and obstacle avoidance.

[0068] Based on a lidar, the obstacles in the surrounding environment of the multi-legged robot are monitored and collected in real time to determine the distance information of the nearby objects of the multi-legged robot for high-precision distance measurement;

[0069] Based on an inertial measurement unit, the acceleration, angular velocity, and attitude angle of the multi-legged robot are monitored and collected in real time to obtain the attitude motion information of the multi-legged robot to assist in the positioning and balance control of the multi-legged robot;

[0070] Among them, based on the environmental vision information, positioning position information, distance information of nearby objects, and attitude and motion information of the multi-legged robot, the real-time motion data of the multi-legged robot is determined.

[0071] In this embodiment, processing the real-time motion data of the multi-legged robot includes:

[0072] Clean the real-time motion data of the multi-legged robot to remove the noise data that is useless for the motion planning of the multi-legged robot;

[0073] Check the real-time motion data of the multi-legged robot to identify duplicate values, missing values, and outliers in the real-time motion data of the multi-legged robot;

[0074] Remove the duplicate values, missing values, and outliers that are useless for the motion planning of the multi-legged robot in the real-time motion data of the multi-legged robot, fill the missing values that are useful for the motion planning of the multi-legged robot in the real-time motion data of the multi-legged robot, and replace the outliers that are useful for the motion planning of the multi-legged robot.

[0075] It should be noted that by cleaning the real-time motion data of the multi-legged robot, the noise data and the duplicate values, missing values, and outliers that are useless for the motion planning of the multi-legged robot can be removed, and the missing values that are useful for the motion planning of the multi-legged robot can be filled and the outliers that are useful for the motion planning of the multi-legged robot can be replaced, which can improve the subsequent processing accuracy and processing speed of the real-time motion data of the multi-legged robot and improve the data quality.

[0076] In this embodiment, processing the real-time motion data of the multi-legged robot further includes:

[0077] Normalize the real-time motion data of the multi-legged robot to convert the real-time motion data of the multi-legged robot into a unified data format, remove the dimensional differences in the real-time motion data of the multi-legged robot, and determine the standardized real-time motion data of the multi-legged robot;

[0078] Extract features from the real-time motion data of the multi-legged robot to extract the features that are useful for the motion planning of the multi-legged robot, and determine the motion feature data of the multi-legged robot.

[0079] Construct a motion planning decision model for the multi-legged robot to analyze the motion feature data of the multi-legged robot, decide the optimal motion planning of the multi-legged robot, determine the motion planning decision scheme of the multi-legged robot, plan the motion of the multi-legged robot, control the gait and actions of the multi-legged robot, and form a closed-loop control of the motion planning of the multi-legged robot in real-time monitoring.

[0080] In this embodiment, constructing a motion planning decision model for the multi-legged robot includes:

[0081] According to the motion planning requirements of the multi-legged robot, collect the motion historical data of the multi-legged robot, and divide the collected motion historical data of the multi-legged robot to determine the training set and the test set;

[0082] Based on deep learning technology, use the training set to train the deep learning model, so that the deep learning model simulates the cognitive mechanism of the human brain and autonomously learns and adapts to complex environments, and optimize the motion planning decision-making scheme to determine the multi-legged robot motion planning decision-making model based on brain-like decision-making;

[0083] Use the test set to test the multi-legged robot motion planning decision-making model based on brain-like decision-making, and evaluate the decision-making performance of the multi-legged robot motion planning decision-making model based on brain-like decision-making to determine the best multi-legged robot motion planning decision-making model.

[0084] In this embodiment, determining the best multi-legged robot motion planning decision-making model includes:

[0085] Evaluate whether the multi-legged robot motion planning decision-making model based on brain-like decision-making can achieve the effect of multi-legged robot motion planning decision-making based on accuracy, recall rate, and F1 score;

[0086] When the multi-legged robot motion planning decision-making model based on brain-like decision-making cannot achieve the effect of multi-legged robot motion planning decision-making, adjust the parameters of the multi-legged robot motion planning decision-making model based on brain-like decision-making, and continuously iterate and optimize the multi-legged robot motion planning decision-making model based on brain-like decision-making until the multi-legged robot motion planning decision-making model based on brain-like decision-making can achieve the effect of multi-legged robot motion planning decision-making, and determine the best multi-legged robot motion planning decision-making model.

[0087] In this embodiment, analyzing the multi-legged robot motion feature data to determine the multi-legged robot motion planning decision-making scheme includes:

[0088] Obtain the best multi-legged robot motion planning decision-making model and deploy the best multi-legged robot motion planning decision-making model in the actual multi-legged robot motion planning environment;

[0089] Input the multi-legged robot motion feature data into the best multi-legged robot motion planning decision-making model, analyze the multi-legged robot motion feature data according to the best multi-legged robot motion planning decision-making model, and make a decision on the optimal multi-legged robot motion planning to determine the multi-legged robot motion planning decision-making scheme;

[0090] Plan the movement of a multi-legged robot according to the movement planning and decision-making scheme of the multi-legged robot, control the gait and actions of the multi-legged robot, plan the movement path of the multi-legged robot, so that the multi-legged robot walks smoothly and can avoid obstacles and reach the target position in a complex environment.

[0091] It should be noted that controlling the gait and actions of the multi-legged robot includes path planning, gait generation, balance control and obstacle avoidance control; among them, path planning is to generate the optimal path from the starting point to the target point based on the environmental map and the movement planning and decision-making scheme of the multi-legged robot; gait generation is to generate suitable gaits according to the terrain and task requirements, such as walking, running, crawling, etc.; balance control is to ensure that the multi-legged robot maintains balance during movement through sensor feedback and real-time control; obstacle avoidance control is to detect obstacles in real time and adjust the movement trajectory to avoid collisions.

[0092] Specifically, determine whether there is an abnormal walking state for the gait parameters during the walking process of the multi-legged robot, and when there is an abnormal walking state, adjust the movement path, including:

[0093] Collect the gait parameters during the walking process of the multi-legged robot in real time, where the gait parameters include the mechanical vibration frequency and the center of gravity deviation value during the robot's walking process;

[0094] Compare the mechanical vibration frequency with a preset mechanical vibration frequency reference value;

[0095] When the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value, then retrieve the center of gravity position at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value;

[0096] Obtain the center of gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value according to the difference degree between the center of gravity position at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the theoretical center of gravity position;

[0097] Utilize the center of gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value;

[0098] Determine whether the current multi-legged robot has an abnormal walking state according to the center of gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the center of gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value;

[0099] When there is an abnormal walking state, adjust the movement path.

[0100] The technical effects of the above technical solution are as follows: By collecting the gait parameters of the multi-legged robot in real time (such as mechanical vibration frequency and center-of-gravity deviation value), the dynamic changes of the robot during walking can be quickly captured. The real-time nature ensures that the system can promptly respond to possible abnormal walking states, thereby improving the response speed and accuracy of the overall system. By comparing the actual mechanical vibration frequency with the preset reference value of the mechanical vibration frequency, the system can accurately determine whether the robot is in an abnormal vibration state. Combining the monitoring of the center-of-gravity deviation value further enhances the system's ability to detect abnormal walking states, because the center-of-gravity deviation is often a direct manifestation of unstable walking. Using the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset value and the previous center-of-gravity deviation values, the system can comprehensively judge whether there is an abnormal walking state in the robot. Once it is determined to be abnormal, the system can immediately adjust the movement path, thereby avoiding safety problems such as collisions or falls that may be caused by unstable walking. This technical solution enables the multi-legged robot to adapt to different walking environments and conditions through real-time monitoring and dynamic path adjustment. Even when encountering uneven ground, obstacles, or other uncertain factors, the system can maintain the stability and safety of the robot by adjusting the path. By accurately monitoring and adjusting the walking state of the robot, this technical solution helps to extend the service life of the robot and reduce mechanical wear and failures caused by unstable walking. At the same time, by optimizing the movement path, the system can also improve the walking efficiency and energy utilization rate of the robot.

[0101] Specifically, determining whether the current multi-legged robot has an abnormal walking state based on the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset reference value of the mechanical vibration frequency and the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the preset reference value of the mechanical vibration frequency includes:

[0102] Retrieving the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset reference value of the mechanical vibration frequency;

[0103] Comparing the center-of-gravity deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset reference value of the mechanical vibration frequency with the preset reference value of the center-of-gravity deviation value;

[0104] Selecting the center-of-gravity deviation values that are not lower than the preset reference value of the center-of-gravity deviation value;

[0105] Retrieving the mechanical vibration frequency of the multi-legged robot corresponding to the moment when the center-of-gravity deviation value is not lower than the preset reference value of the center-of-gravity deviation value;

[0106] Obtaining the standard deviation of the mechanical vibration frequency based on the mechanical vibration frequency of the multi-legged robot corresponding to the moment when the center-of-gravity deviation value is not lower than the preset reference value of the center-of-gravity deviation value;

[0107] Obtain an abnormal evaluation coefficient by combining the center-of-gravity deviation value not less than the reference value of the preset center-of-gravity deviation value and the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the reference value of the preset mechanical vibration frequency with the standard deviation of the mechanical vibration frequency;

[0108] Among them, the abnormal evaluation coefficient is obtained through the following formula:

[0109]

[0110] Among them, Y represents the abnormal evaluation coefficient; Y0 represents the preset coefficient reference value; n represents the number of center-of-gravity deviation values not less than the reference value of the preset center-of-gravity deviation value; Z i represents the value of the i-th center-of-gravity deviation value not less than the reference value of the preset center-of-gravity deviation value; Z c represents the reference value of the preset center-of-gravity deviation value; Z represents the center-of-gravity deviation value at the moment when the mechanical vibration frequency exceeds the reference value of the preset mechanical vibration frequency; f represents the mechanical vibration frequency of the robot exceeding the reference value of the preset mechanical vibration frequency; f b represents the standard deviation of the mechanical vibration frequency;

[0111] Compare the abnormal evaluation coefficient with the preset coefficient threshold;

[0112] When the abnormal evaluation coefficient exceeds the preset coefficient threshold, it is determined that the walking state of the multi-legged robot is abnormal.

[0113] The technical effect of the above technical solution is: In the above technical solution, The numerator part of the above formula calculates the difference between the center-of-gravity deviation value at a specific moment (when the mechanical vibration frequency exceeds the preset reference value) and the center-of-gravity deviation value Zi of the i-th center-of-gravity deviation value not less than the preset center-of-gravity deviation value reference value Zc, reflecting the difference in the center-of-gravity deviation value under different states; multiplying by the mechanical vibration frequency f at this time associates the center-of-gravity deviation difference with the vibration frequency and considers the influence weight of the vibration frequency on this deviation difference. The denominator part reflects the degree of deviation of the current vibration frequency from its fluctuation range. The whole denominator is used to normalize the calculation result of the numerator.

[0114] Through the above calculation, an accumulated value comprehensively reflecting the relationship between the center-of-gravity deviation value and the mechanical vibration frequency is obtained, quantifying the comprehensive influence degree of the vibration frequency on the system state under different center-of-gravity deviation states. Then, taking the average of the previous accumulated value is to eliminate the influence of possible individual abnormal values or fluctuations in the n data, making the calculation result more representative and stable, and obtaining a quantitative index that comprehensively reflects the relationship between the center-of-gravity deviation and the vibration frequency in an average sense. It can effectively provide a more stable value representing the overall situation for subsequent calculation of the abnormal evaluation coefficient in combination with other factors.

[0115] On the other hand, by comprehensively considering two key gait parameters, namely the mechanical vibration frequency and the center of gravity deviation value, as well as their historical data and standard deviation, this technical solution can more accurately determine whether there is an abnormality in the walking state of the multi-legged robot. By introducing an abnormal evaluation coefficient Y and conducting a comprehensive evaluation in combination with multiple factors (such as the number and value of the center of gravity deviation values, the mechanical vibration frequency, and its standard deviation), the accuracy and reliability of the judgment are improved. By real-time monitoring and comparing the mechanical vibration frequency and the center of gravity deviation value, this technical solution can provide early warnings before the occurrence of an abnormal walking state. It can timely detect and handle potential walking problems, avoid the further development and expansion of faults, thereby ensuring the safety and stability of the robot. This technical solution can dynamically adjust the judgment criteria according to the actual walking situation of the multi-legged robot, such as the preset reference value of the center of gravity deviation value and the preset coefficient threshold. This adaptability enables the technical solution to be applicable to multi-legged robots of different models, different loads, and different walking environments, improving the robustness and versatility of the system. By quickly calculating the abnormal evaluation coefficient and comparing it with the preset coefficient threshold, this technical solution can quickly determine whether the walking state of the multi-legged robot is abnormal. This improves the efficiency of fault detection and reduces the potential risks caused by judgment delays. Once it is determined that there is an abnormality in the walking state of the multi-legged robot, this technical solution can provide decision-making support for subsequent path adjustment or fault handling. It can ensure that the robot can take effective measures in a timely manner when encountering walking problems, maintaining the stability and safety of walking. By long-term monitoring and analysis of the walking state data of the multi-legged robot, this technical solution can provide strong support for the performance optimization and continuous improvement of the robot. It can enhance the overall performance of the robot, extend its service life, and reduce the maintenance cost.

[0116] In this embodiment, real-time monitoring forms a closed-loop control for the motion planning of the multi-legged robot, including:

[0117] Real-time monitoring of the motion of the multi-legged robot and real-time feedback of the monitoring data to the motion planning decision-making model of the multi-legged robot to form a closed-loop control for the motion planning of the multi-legged robot, and then dynamically adjust and optimize the motion planning decision-making scheme of the multi-legged robot, and conduct closed-loop management of the motion planning of the multi-legged robot.

[0118] In summary, by collecting the environmental visual information, positioning location information, distance information of nearby objects, and attitude motion information of the multi-legged robot, the real-time motion data of the multi-legged robot is determined. After processing the real-time motion data of the multi-legged robot, the motion feature data of the multi-legged robot is determined. According to the motion planning requirements of the multi-legged robot, a motion planning decision model of the multi-legged robot is constructed to analyze the motion feature data of the multi-legged robot, and the optimal motion planning of the multi-legged robot is decided, and the motion planning decision scheme of the multi-legged robot is determined. Then, according to the motion planning decision scheme of the multi-legged robot, the motion of the multi-legged robot is planned, the gait and actions of the multi-legged robot are controlled, and the motion of the multi-legged robot is monitored in real time to form a closed-loop control of the motion planning of the multi-legged robot. It can effectively plan the motion of the multi-legged robot based on brain-inspired decision-making and visual positioning, improve the autonomous navigation and motion planning decision-making ability of the multi-legged robot in complex environments, make the motion effect of the multi-legged robot good, and can be applied to disaster rescue scenarios to perform search and rescue tasks at disaster sites and adapt to complex terrains; it can be applied to field exploration scenarios to perform exploration tasks in unknown environments and collect environmental data; it can be applied to industrial inspection scenarios to perform equipment inspection tasks in factories or dangerous environments; it can be applied to military reconnaissance scenarios to perform reconnaissance tasks in battlefield environments and adapt to various terrains.

[0119] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0120] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-legged robot motion planning method based on brain-like decision-making and visual positioning, characterized in that: include: Collecting real-time data of multi-legged robot motion and processing it to determine the motion characteristic data of the multi-legged robot; Construct a multi-legged robot motion planning decision model to analyze the multi-legged robot motion characteristic data, decide on the optimal multi-legged robot motion planning, and determine the multi-legged robot motion planning decision plan; Plan the motion of the multi-legged robot, control the gait and movement of the multi-legged robot, and monitor in real time to form a closed-loop control of the multi-legged robot motion planning; Construct a multi-legged robot motion planning decision model, including: According to the requirements of multi-legged robot motion planning, the multi-legged robot motion history data is collected, and the collected multi-legged robot motion history data is divided to determine the training set and the test set; Based on deep learning technology, the deep learning model is trained with a training set, so that the deep learning model can simulate the cognitive mechanism of the human brain and autonomously learn and adapt to complex environments, optimize the motion planning decision-making scheme, and determine the motion planning decision-making model of a multi-legged robot based on brain-like decision-making; The test set is used to test the multi-legged robot motion planning decision model based on brain-like decision-making, and the decision-making performance of the multi-legged robot motion planning decision model based on brain-like decision-making is evaluated to determine the best multi-legged robot motion planning decision model; Analyze the motion characteristic data of the multi-legged robot and determine the motion planning decision-making plan of the multi-legged robot, including: Obtain the best multi-legged robot motion planning decision model, and deploy the best multi-legged robot motion planning decision model in the actual multi-legged robot motion planning environment; Inputting the multi-legged robot motion characteristic data into the optimal multi-legged robot motion planning decision model, analyzing the multi-legged robot motion characteristic data according to the optimal multi-legged robot motion planning decision model, determining the optimal multi-legged robot motion planning, and determining the multi-legged robot motion planning decision plan; According to the multi-legged robot motion planning decision plan, the multi-legged robot motion is planned, the gait and movement of the multi-legged robot are controlled, and the motion path of the multi-legged robot is planned, so that the multi-legged robot can walk smoothly and avoid obstacles in complex environments and reach the target position.

2. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 1, characterized in that: Collect real-time data of multi-legged robot movement, including: Based on the visual sensor, the surrounding environment of the multi-legged robot is monitored and collected in real time to determine the environmental visual information of the multi-legged robot; Based on the environmental visual data of the multi-legged robot and combined with the real-time positioning and map building technology, the multi-legged robot is visually positioned and the environmental map is built to determine the positioning position information of the multi-legged robot; Based on the laser radar, the obstacles in the environment around the multi-legged robot are monitored and collected in real time to determine the distance information of the close-range objects of the multi-legged robot; Based on the inertial measurement unit, the acceleration, angular velocity and attitude angle of the multi-legged robot are monitored and collected in real time to obtain the attitude motion information of the multi-legged robot; The real-time motion data of the multi-legged robot is determined based on the environmental visual information, positioning position information, close-range object distance information and posture motion information of the multi-legged robot.

3. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 1, characterized in that: Processing of real-time data of multi-legged robot motion, including: Clean the real-time data of the multi-legged robot's motion and remove the noise data in the real-time data of the multi-legged robot's motion that is useless for the multi-legged robot's motion planning; Check the real-time data of multi-legged robot motion and identify duplicate values, missing values ​​and abnormal values ​​in the real-time data of multi-legged robot motion; The duplicate values, missing values ​​and abnormal values ​​in the real-time motion data of the multi-legged robot that are useless for the motion planning of the multi-legged robot are removed, the missing values ​​in the real-time motion data of the multi-legged robot that are useful for the motion planning of the multi-legged robot are filled, and the abnormal values ​​that are useful for the motion planning of the multi-legged robot are replaced.

4. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 3, characterized in that: Processing of real-time data of multi-legged robot movement also includes: Normalizing the real-time motion data of the multi-legged robot, converting the real-time motion data of the multi-legged robot into a unified data format, removing the dimensional differences in the real-time motion data of the multi-legged robot, and determining standardized real-time motion data of the multi-legged robot; Feature extraction is performed on the real-time motion data of the multi-legged robot, features useful for the motion planning of the multi-legged robot are extracted from the real-time motion data of the multi-legged robot, and the motion feature data of the multi-legged robot is determined.

5. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 1, characterized in that: Determine the best multi-legged robot motion planning decision model, including: Based on the accuracy, recall and F1 score, the multi-legged robot motion planning decision model based on brain-like decision-making is evaluated to see whether it can achieve the effect of multi-legged robot motion planning decision-making; When the multi-legged robot motion planning decision model based on brain-like decision-making cannot achieve the effect of the multi-legged robot motion planning decision-making, the parameters of the multi-legged robot motion planning decision model based on brain-like decision-making are adjusted, and the multi-legged robot motion planning decision model based on brain-like decision-making is continuously iterated and optimized until the multi-legged robot motion planning decision model based on brain-like decision-making can achieve the effect of the multi-legged robot motion planning decision-making, and the optimal multi-legged robot motion planning decision model is determined.

6. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 1, characterized in that: The gait parameters of the multi-legged robot during walking are used to determine whether there is an abnormal walking state, and when there is an abnormal walking state, the motion path is adjusted, including: Collecting gait parameters of the multi-legged robot during walking in real time, wherein the gait parameters include mechanical vibration frequency and center of gravity deviation value during the robot's walking process; comparing the mechanical vibration frequency with a preset mechanical vibration frequency reference value; When the mechanical vibration frequency exceeds a preset mechanical vibration frequency reference value, the center of gravity position at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value is retrieved; Obtaining a gravity center deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value according to a difference between the gravity center position at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the theoretical gravity center position; Utilizing the gravity center deviation value of the robot before the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value; Determining whether the current multi-legged robot has an abnormal walking state according to the gravity center deviation value of the robot before the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the gravity center deviation value when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value; When there is an abnormality in the walking state, the motion path is adjusted.

7. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 6, characterized in that: Determining whether the current multi-legged robot has an abnormal walking state according to the gravity center deviation value of the robot before the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value and the gravity center deviation value when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value includes: Retrieving the gravity center deviation value of the robot before the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value; Comparing the center of gravity deviation value of the robot before the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value with a preset center of gravity deviation value reference value; Screening out a center of gravity deviation value that is not less than a preset center of gravity deviation value reference value; Retrieving the mechanical vibration frequency of the multi-legged robot corresponding to the moment when the center of gravity deviation value is not less than a preset center of gravity deviation value reference value; Obtaining a standard deviation of the mechanical vibration frequency according to the mechanical vibration frequency of the multi-legged robot corresponding to the moment when the gravity center deviation value is not less than a preset gravity center deviation value reference value; Obtaining an abnormality evaluation coefficient using the gravity center deviation value that is not less than the preset gravity center deviation value reference value and the gravity center deviation value at the moment when the mechanical vibration frequency exceeds the preset mechanical vibration frequency reference value in combination with the mechanical vibration frequency standard deviation; Comparing the abnormal evaluation coefficient with a preset coefficient threshold; When the abnormality evaluation coefficient exceeds a preset coefficient threshold, it is determined that there is an abnormality in the walking state of the multi-legged robot.

8. The multi-legged robot motion planning method based on brain-like decision-making and visual positioning as claimed in claim 1, characterized in that: Real-time monitoring forms a closed-loop control of multi-legged robot motion planning, including: The motion status of the multi-legged robot is monitored in real time, and the monitoring data is fed back to the multi-legged robot motion planning decision model in real time to form a closed-loop control of the multi-legged robot motion planning, and then the multi-legged robot motion planning decision scheme is dynamically adjusted and optimized to perform closed-loop management of the multi-legged robot motion planning.

Images