Method and system for evaluating real-time control performance of humanoid robot

Abstract

The invention relates to the technical field of humanoid robot control, and discloses a humanoid robot real-time control performance evaluation method, which comprises the steps of calculating a control output quantity of humanoid robot obstacle crossing gait control in a set standard state; obstacle crossing control is carried out on the humanoid robot under different terrains and different obstacles, so that a control deviation standard value is obtained; setting a plurality of centroid heights when the humanoid robot crosses different obstacles, and taking the centroid heights as standard fluctuation ranges of the centroid heights; and establishing a control comparison table, monitoring the control deviation and the centroid height of the humanoid robot during obstacle crossing in real time, and comparing the control deviation and the centroid height with corresponding standard values in the control comparison table to determine the real-time control performance of the humanoid robot. The control performance of the humanoid robot during obstacle crossing is evaluated, the stability problem of the humanoid robot during obstacle crossing can be known in time, control of the humanoid robot can be adjusted in time, and the gait stability of the humanoid robot during obstacle crossing is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of humanoid robot control, and in particular to a method and system for evaluating the real-time control performance of a humanoid robot. Background Art

[0002] Humanoid mobile robots are complex systems with multiple degrees of freedom and nonlinear constraints. Controlling them to avoid or traverse obstacles is a challenging task. The core difficulty lies in the fact that when a humanoid robot traverses an obstacle, its center of gravity tilts significantly forward, allowing it to gain more space to traverse. This, however, conflicts with the constraints that maintain the robot's stability. Furthermore, the robot's upper body has poor balance, making it more prone to instability and falls when traversing obstacles. This results in varying control performance for humanoid robots in traversing obstacles.

[0003] The gait control performance of a humanoid robot when crossing obstacles affects the gait stability of the humanoid robot when crossing obstacles. In the existing technology, there is no method to evaluate the real-time control performance of a humanoid robot, and it is impossible to make targeted adjustments to the gait of the humanoid robot when crossing obstacles to make the gait of the humanoid robot more stable. Summary of the Invention

[0004] The present invention provides a real-time control performance evaluation method and system for a humanoid robot, which evaluates the control performance of a humanoid robot when crossing an obstacle. The method can timely understand the stability problems of the humanoid robot when crossing an obstacle, so as to timely adjust the control of the humanoid robot and ensure the stability of the gait of the humanoid robot when crossing the obstacle.

[0005] The present invention provides a method for evaluating the real-time control performance of a humanoid robot, comprising:

[0006] calculating a control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot;

[0007] performing obstacle crossing control on the humanoid robot under different terrains and obstacles according to the control output to obtain a standard value of control deviation;

[0008] Measuring by a tilt platform method to set multiple center of mass heights of the humanoid robot when it faces different obstacles and surmounts them, and setting the multiple center of mass heights as a standard fluctuation range of the center of mass height;

[0009] A unique mapping relationship is formed between the control deviation standard value, the center of mass height standard fluctuation range and the corresponding terrain and obstacles, and a control comparison table is established;

[0010] monitoring the control deviation and center of mass height of the humanoid robot in real time when crossing an obstacle, and comparing the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles;

[0011] The real-time control performance of the humanoid robot is determined according to the comparison result; wherein the real-time control performance includes good, medium, and poor.

[0012] Furthermore, the step of calculating the control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot includes:

[0013] Divide the humanoid robot into six parts: a torso, a head, a left upper limb, a right upper limb, a left lower limb, and a right lower limb, and create a part detector for the humanoid robot;

[0014] Calculating correlation indices of adjacent parts of the six parts of the humanoid robot, and obtaining posture information of the humanoid robot according to the correlation indices;

[0015] The posture information of the humanoid robot and the image information of the obstacle are input into a neural network model based on feedforward compensation to perform gait control, and a control output of the obstacle-crossing gait control of the humanoid robot is obtained.

[0016] Furthermore, the step of dividing the humanoid robot into six parts, namely, a torso, a head, a left upper limb, a right upper limb, a left lower limb, and a right lower limb, and creating a humanoid robot part detector comprises:

[0017] Crop the selected part from the training image, rotate it to a vertical position, and collect multiple images of the robot body as negative examples;

[0018] Calculate the directional gradient histogram features of each sample image and obtain a vector Z. Vector Z plus the positive or negative label represents the corresponding sample image;

[0019] The obtained vector is introduced into the AdaBoost algorithm to train a strong classifier, which is recorded as:

[0020]

[0021] Among them, S(·) is the sigmoid function, h i (·) is the function for calculating the length of the i-th rectangle.

[0022] Furthermore, the step of calculating correlation indices of adjacent parts among the six parts of the humanoid robot and obtaining posture information of the humanoid robot according to the correlation indices includes:

[0023] Assume that the connected edge in the connected part is g, and the connected parts are connected to each other at point k. Then the distance between the center point of edge g and point k is:

[0024]

[0025] Among them, x g 、y g is the horizontal and vertical coordinates of edge g; x k 、y k is the horizontal and vertical coordinates of vertex k;

[0026] The correlation index is expressed as:

[0027] o d =p (―1×(d―n)) / (N―n)

[0028] Among them, p is the association constant, n is the minimum d value in all candidate areas, and N is the maximum d value in all candidate areas. The smaller the d value, the higher the association degree. d The larger the value of , the higher the correlation degree of the humanoid robot posture information.

[0029] Furthermore, the step of inputting the posture information of the humanoid robot and the image information of the obstacle into a neural network model based on feedforward compensation to perform gait control and obtain a control output of the obstacle-crossing gait control of the humanoid robot includes:

[0030] The image information of the obstacle is obtained by using a camera. The image information is described using a coordinate system, and the obstacle projection orientation is expressed as:

[0031]

[0032] Among them, u and w are the horizontal and vertical reference point orientations;

[0033] Setting I (1) , O (1) are the input value and output value of the first layer of the neural network model based on feedforward compensation. In the input layer, the acceleration sensor measurement signal f of the humanoid robot and the posture information are used as input and substituted into the input end of the network controller. At the same time, the output value f is i Passed to the next layer of the network, the following corresponding relationship exists at this time, namely:

[0034] O (1) =I (1) =f

[0035] Inside the hidden layer, the sigmoid function is used to obtain the activation function of each node input:

[0036]

[0037] Among them, s i is the function weight value under the i-th node;

[0038] In the output layer, the information obtained by the hidden layer is linearly calculated, and the calculation results are used as weights to obtain:

[0039]

[0040] Where R is the total number of training times;

[0041] Set the range of the objective function F(x), recorded as:

[0042]

[0043] Among them, z is the output control quantity of the control system, z n is the compensation of the neural network discriminator;

[0044] In order to minimize the objective function value, the compensation amount of the control cycle is set to t c , the output of the obstacle crossing gait controller is t o , the current control quantity t(j) of the control system is the sum of the above two. The sum value can be calculated to complete the training of the neural network model and the intelligent obstacle-crossing gait control task of the humanoid robot. The calculation formula of the control output sum is:

[0045] t(j)=t c (j)+t o (j)

[0046] Among them, j is the time required for current control.

[0047] Furthermore, the step of monitoring the control deviation and center of mass height of the humanoid robot in real time when crossing an obstacle, and comparing the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles, includes:

[0048] monitoring in real time the control deviation and center of mass height of the humanoid robot when crossing an obstacle, and obtaining the terrain and obstacle corresponding to the current control deviation, and obtaining the obstacle corresponding to the current center of mass height;

[0049] Comparing the control deviation with the standard value of the control deviation under the corresponding terrain and obstacles;

[0050] The multiple center-of-mass heights of the humanoid robot when crossing an obstacle are used to form a current center-of-mass height fluctuation range, and the current center-of-mass height fluctuation range is compared with the center-of-mass height standard fluctuation range according to a set rule, where the set rule is:

[0051] Calculate the difference between the maximum value of the current center of mass height fluctuation range and the maximum value of the center of mass height standard fluctuation range, and compare it with the set threshold;

[0052] The highest value and the lowest value of the current center of mass height fluctuation range are compared with the highest value and the lowest value of the center of mass height standard fluctuation range.

[0053] Furthermore, the step of determining the real-time control performance of the humanoid robot according to the comparison result includes:

[0054] When the control deviation is less than a set value and the center of mass height fluctuation is normal, it is determined that the control performance of the humanoid robot is good;

[0055] When the control deviation is less than a set value and the center of mass height fluctuates abnormally, it is determined that the control performance of the humanoid robot is medium;

[0056] When the control deviation is greater than or equal to a set value, it is determined that the control performance of the humanoid robot is poor.

[0057] Furthermore, the abnormal fluctuation of the center of mass height is as follows:

[0058] Calculating the difference between the highest value of the current center of mass height fluctuation range and the highest value of the center of mass height standard fluctuation range, and when the difference is greater than or equal to a set threshold, determining that the center of mass height fluctuation is abnormal;

[0059] When the maximum value of the current center of mass height fluctuation range is higher than the maximum value of the center of mass height standard fluctuation range, and the minimum value of the current center of mass height fluctuation range is lower than the minimum value of the center of mass height standard fluctuation range, it is determined that the center of mass height fluctuation is abnormal.

[0060] The present invention also provides a humanoid robot real-time control performance evaluation system, comprising:

[0061] A calculation module, configured to calculate a control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot;

[0062] an acquisition module, configured to perform obstacle-crossing control on the humanoid robot under different terrains and obstacles according to the control output, so as to obtain a control deviation standard value;

[0063] a setting module, configured to measure and set multiple center of mass heights of the humanoid robot when facing different obstacles and overcoming them by using a tilt platform method, and to set the multiple center of mass heights as a standard fluctuation range of the center of mass height;

[0064] Establishing a module for forming a unique mapping relationship between the control deviation standard value, the center of mass height standard fluctuation range and the corresponding terrain and obstacles, and establishing a control comparison table;

[0065] a comparison module, configured to monitor in real time the control deviation and center of mass height of the humanoid robot when it crosses an obstacle, and compare the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles;

[0066] A determination module is used to determine the real-time control performance of the humanoid robot according to the comparison result; wherein the real-time control performance includes good, medium, and poor.

[0067] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0068] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0069] The beneficial effects of the present invention are:

[0070] The present invention first calculates the control output of the humanoid robot's obstacle-crossing gait control under a standard state based on the humanoid robot's posture information. The robot then performs obstacle-crossing control, obtaining a control deviation as a standard value and setting it as a standard fluctuation range for the center of mass height. A unique mapping relationship is then established between the control deviation, center of mass height, and the corresponding terrain and obstacles to establish a control comparison table. Finally, the control deviation and center of mass height of the humanoid robot during obstacle crossing are monitored in real time and compared with the corresponding standard values ​​in the control comparison table. The comparison results are used to determine the real-time control performance of the humanoid robot, which can be classified as good, medium, or poor. By evaluating the control performance of the humanoid robot during obstacle crossing, the stability issues of the humanoid robot during obstacle crossing can be promptly identified, allowing for timely adjustments to the humanoid robot's control to ensure a stable gait during obstacle crossing. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 Schematic diagram of a method flow according to an embodiment of the present invention.

[0072] Figure 2 FIG. 1 is a schematic diagram of the device structure according to an embodiment of the present invention.

[0073] Figure 3 Schematic diagram of the internal structure of a computer device according to an embodiment of the present invention.

[0074] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0075] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0076] like Figure 1 As shown, the present invention provides a method for evaluating the real-time control performance of a humanoid robot, comprising:

[0077] S1. Calculating a control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot.

[0078] S101. Divide the humanoid robot into six parts: a torso, a head, a left upper limb, a right upper limb, a left lower limb, and a right lower limb, and create a humanoid robot part detector.

[0079] Introducing posture information into obstacle-crossing gait control allows for more accurate control of the robot's movement posture and precise obstacle crossing. The bipedal mobile robot is divided into six parts: the torso, head, left upper limb, right upper limb, left lower limb, and right lower limb. A robot part detector is created to locate the corresponding robot parts in an image. The following process is used to create a robot part detector:

[0080] 1>Crop the selected part from the training image, rotate it to a vertical position, and collect multiple images of the robot body as negative examples;

[0081] 2> Calculate the directional gradient histogram features of each sample image and obtain a vector Z. Vector Z plus the positive or negative label represents the corresponding sample image;

[0082] 3>Introduce the obtained vector into the AdaBoost algorithm to train a strong classifier, which is recorded as:

[0083]

[0084] Among them, S(·) is the sigmoid function, h i (·) is the function for calculating the length of the i-th rectangle.

[0085] S102, calculating the correlation index of adjacent parts among the six parts of the humanoid robot, and obtaining the posture information of the humanoid robot according to the correlation index. Specifically including:

[0086] Correlation is a key indicator for obtaining robot posture information. Taking the distance relationship between the left upper limb and the torso as an example, assuming the torso's orientation is known, we only need to calculate the current orientation of the left upper limb. Assuming the edge connecting the left upper limb and the torso is g, and the two are connected at the torso vertex k, the distance between the center point of edge g and the torso vertex k is:

[0087]

[0088] Among them, x g 、y g is the horizontal and vertical coordinates of edge g; x k 、y k is the horizontal and vertical coordinates of vertex k;

[0089] The correlation index is expressed as:

[0090] o d =p (―1×(d―n)) / (N―n)

[0091] Among them, p is the association constant, n is the minimum d value in all candidate areas, and N is the maximum d value in all candidate areas. The smaller the d value, the higher the association degree. d The larger the value of , the higher the correlation degree of the humanoid robot posture information.

[0092] Bone density can reveal the position and orientation of every part of a robot, and is a core identification feature in human-machine interaction. By acquiring bone data, the system evaluates the robot's posture and calculates the bone density of the candidate area as follows:

[0093] q j =r1q1+r2q2

[0094] Among them, q1 is the number of skeleton points, q2 is the center level of the skeleton points, r1 and r2 are both calculation weights, and the sum of the two is equal to 1.

[0095] S103: Inputting the posture information of the humanoid robot and the image information of the obstacle into a neural network model based on feedforward compensation to perform gait control, and obtaining a control output of the obstacle-crossing gait control of the humanoid robot.

[0096] The image information of the obstacle is obtained by using a camera. The image information is described using a coordinate system, and the obstacle projection orientation is expressed as:

[0097]

[0098] Among them, u and w are the horizontal and vertical reference point orientations;

[0099] When a humanoid mobile robot traverses obstacles, it uses a neural network with feedforward compensation capabilities for gait control. This model addresses the adverse effects of bending the robot's wires, creating a feedforward control model that establishes a compensation model between the deviation and the control variable. The neural network feedforward compensation controller and the neural network discriminator have identical architectures, consisting of input, hidden, and output layers. The difference lies in the parameters introduced at the input and output levels.

[0100] Setting I( 1) 、O(1) are the input value and output value of the first layer of the neural network model based on feedforward compensation. In the input layer, the acceleration sensor measurement signal f of the humanoid robot and the posture information are used as input and substituted into the input end of the network controller. At the same time, the output value f is i Passed to the next layer of the network, the following corresponding relationship exists at this time, namely:

[0101] O (1) =I (1) =f

[0102] Inside the hidden layer, the sigmoid function is used to obtain the activation function of each node input:

[0103]

[0104] Among them, s i is the function weight value under the i-th node;

[0105] In the output layer, the information obtained by the hidden layer is linearly calculated, and the calculation results are used as weights to obtain:

[0106]

[0107] Where R is the total number of training times;

[0108] The gradient descent algorithm is easy to calculate and has strong timeliness. It is used to complete the calculation task in neural network parameter learning. First, set the value range of the objective function F(x), which is expressed as:

[0109]

[0110] Among them, z is the output control quantity of the control system, z n is the compensation of the neural network discriminator;

[0111] The fundamental purpose of neural network learning is to minimize the objective function value and set the compensation amount of the control cycle to t c , the output of the obstacle crossing gait controller is t o , the current control quantity t(j) of the control system is the sum of the above two. The sum value can be calculated to complete the training of the neural network model and the intelligent obstacle-crossing gait control task of the humanoid robot. The calculation formula of the control output sum is:

[0112] t(j)=t c (j)+t o (j)

[0113] Among them, j is the time required for current control.

[0114] S2. Performing obstacle-crossing control on the humanoid robot under different terrains and obstacles according to the control output, and obtaining a control deviation as a standard value in real time during the obstacle-crossing control.

[0115] S3. Measure and set multiple center of mass heights when the humanoid robot faces different obstacles and overcomes them by using the tilt platform method, and set the multiple center of mass heights as the standard fluctuation range of the center of mass height; that is, take the lowest value to the highest value of the multiple center of mass heights as the standard fluctuation range of the center of mass height.

[0116] S4. Uniquely map the control deviation standard value, center of mass height standard fluctuation range, and the corresponding terrain and obstacles, and establish a control comparison table; that is, one terrain and one obstacle correspond to one control deviation standard value and one center of mass height standard fluctuation range.

[0117] S5. Monitor the control deviation and center of mass height of the humanoid robot in real time when it crosses an obstacle, and compare the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles.

[0118] The specific steps include:

[0119] S501. Monitor the control deviation and center of mass height of the humanoid robot in real time when crossing obstacles, and obtain the terrain and obstacle corresponding to the current control deviation, and obtain the obstacle corresponding to the current center of mass height; that is, when performing obstacle crossing control, record the terrain and obstacles being controlled, and record the control deviation and center of mass height range.

[0120] S502, comparing the control deviation with a standard value of the control deviation under corresponding terrain and obstacles;

[0121] S503: The multiple center-of-mass heights of the humanoid robot when crossing an obstacle are used to form a current center-of-mass height fluctuation range, and the current center-of-mass height fluctuation range is compared with the center-of-mass height standard fluctuation range according to a set rule, where the set rule is:

[0122] S504. Calculate the difference between the maximum value of the current center of mass height fluctuation range and the maximum value of the standard center of mass height fluctuation range, and compare it with the set threshold; when the calculated difference is greater than or equal to the set threshold, it can be determined that the center of mass height fluctuation is abnormal; when the calculated difference is less than the set threshold, it can be determined that the center of mass height fluctuation is normal.

[0123] S505: Compare the highest value and the lowest value of the current center of mass height fluctuation range with the highest value and the lowest value of the center of mass height standard fluctuation range.

[0124] If the maximum value of the current center of mass height fluctuation range is higher than the maximum value of the center of mass height standard fluctuation range, and the minimum value of the current center of mass height fluctuation range is lower than the minimum value of the center of mass height standard fluctuation range, it can be determined that the center of mass height fluctuation is abnormal; in other cases, it can be determined that the center of mass height fluctuation is normal.

[0125] S6. Determine the real-time control performance of the humanoid robot according to the comparison result; wherein the real-time control performance includes good, medium, and poor.

[0126] The specific steps include:

[0127] S601: When the control deviation is less than a set value and the center of mass height fluctuation is normal (as in step S505), it is determined that the control performance of the humanoid robot is good.

[0128] S602: When the control deviation is less than a set value and the center of mass height fluctuates abnormally, it is determined that the control performance of the humanoid robot is medium.

[0129] The abnormal fluctuation of the center of mass height is as follows:

[0130] 1) calculating the difference between the maximum value of the current center of mass height fluctuation range and the maximum value of the center of mass height standard fluctuation range, and when the difference is greater than or equal to a set threshold, determining that the center of mass height fluctuation is abnormal;

[0131] 2) When the maximum value of the current center of mass height fluctuation range is higher than the maximum value of the center of mass height standard fluctuation range, and the minimum value of the current center of mass height fluctuation range is lower than the minimum value of the center of mass height standard fluctuation range, it is determined that the center of mass height fluctuation is abnormal.

[0132] Any situation that does not fall under the aforementioned abnormal fluctuation of the center of mass height can be determined as the normal fluctuation of the center of mass height.

[0133] S603: When the control deviation is greater than or equal to a set value, determine that the control performance of the humanoid robot is poor.

[0134] The present invention first calculates the control output of the humanoid robot's obstacle-crossing gait control under a standard state based on the humanoid robot's posture information. The robot then performs obstacle-crossing control, obtaining a control deviation as a standard value and setting it as a standard fluctuation range for the center of mass height. A unique mapping relationship is then established between the control deviation, center of mass height, and the corresponding terrain and obstacles to establish a control comparison table. Finally, the control deviation and center of mass height of the humanoid robot during obstacle crossing are monitored in real time and compared with the corresponding standard values ​​in the control comparison table. The comparison results are used to determine the real-time control performance of the humanoid robot, which can be classified as good, medium, or poor. By evaluating the control performance of the humanoid robot during obstacle crossing, the stability issues of the humanoid robot during obstacle crossing can be promptly identified, allowing for timely adjustments to the humanoid robot's control to ensure a stable gait during obstacle crossing.

[0135] like Figure 2 As shown, the present invention also provides a humanoid robot real-time control performance evaluation system, comprising:

[0136] A calculation module 1 is used to calculate the control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot;

[0137] an acquisition module 2, configured to perform obstacle crossing control on the humanoid robot under different terrains and different obstacles according to the control output, so as to obtain a control deviation standard value;

[0138] A setting module 3 is configured to measure and set multiple center of mass heights of the humanoid robot when it faces different obstacles and overcomes them by using a tilt platform method, and to set the multiple center of mass heights as a standard fluctuation range of the center of mass height;

[0139] Establishing module 4, for forming a unique mapping relationship between the control deviation standard value, the center of mass height standard fluctuation range and the corresponding terrain and obstacles, and establishing a control comparison table;

[0140] Comparison module 5, for real-time monitoring of the control deviation and center of mass height of the humanoid robot when crossing an obstacle, and comparing the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles;

[0141] The determination module 6 is configured to determine the real-time control performance of the humanoid robot according to the comparison result; wherein the real-time control performance includes good, medium, and poor.

[0142] In one embodiment, the computing module 1 includes:

[0143] a creation unit, configured to divide the humanoid robot into six parts: a torso, a head, a left upper limb, a right upper limb, a left lower limb, and a right lower limb, and to create a part detector for the humanoid robot;

[0144] a correlation index calculation unit, configured to calculate correlation indexes of adjacent parts of the six parts of the humanoid robot, and obtain posture information of the humanoid robot according to the correlation indexes;

[0145] A control unit is used to input the posture information of the humanoid robot and the image information of the obstacle into a neural network model based on feedforward compensation to perform gait control, and obtain a control output of the obstacle-crossing gait control of the humanoid robot.

[0146] In one embodiment, the creating unit includes:

[0147] Crop the selected part from the training image, rotate it to a vertical position, and collect multiple images of the robot body as negative examples;

[0148] Calculate the directional gradient histogram features of each sample image and obtain a vector Z. Vector Z plus the positive or negative label represents the corresponding sample image;

[0149] The obtained vector is introduced into the AdaBoost algorithm to train a strong classifier, which is recorded as:

[0150]

[0151] Among them, S(·) is the sigmoid function, h i (·) is the function for calculating the length of the i-th rectangle.

[0152] In one embodiment, the relevance index calculation unit includes:

[0153] Assume that the connected edge in the connected part is g, and the connected parts are connected to each other at point k. Then the distance between the center point of edge g and point k is:

[0154]

[0155] Among them, x g 、y g is the horizontal and vertical coordinates of edge g; x k 、y k is the horizontal and vertical coordinates of vertex k;

[0156] The correlation index is expressed as:

[0157] o d =p (―1×(d―n)) / (N―n)

[0158] Among them, p is the association constant, n is the minimum d value in all candidate areas, and N is the maximum d value in all candidate areas. The smaller the d value, the higher the association degree. dThe larger the value of , the higher the correlation degree of the humanoid robot posture information.

[0159] In one embodiment, the control unit comprises:

[0160] The image information of the obstacle is obtained by using a camera. The image information is described using a coordinate system, and the obstacle projection orientation is expressed as:

[0161]

[0162] Among them, u and w are the horizontal and vertical reference point orientations;

[0163] Setting I (1) , O (1) are the input value and output value of the first layer of the neural network model based on feedforward compensation. In the input layer, the acceleration sensor measurement signal f of the humanoid robot and the posture information are used as input and substituted into the input end of the network controller. At the same time, the output value f is i Passed to the next layer of the network, the following corresponding relationship exists at this time, namely:

[0164] O (1) =I (1) =f

[0165] Inside the hidden layer, the sigmoid function is used to obtain the activation function of each node input:

[0166]

[0167] Among them, s i is the function weight value under the i-th node;

[0168] In the output layer, the information obtained by the hidden layer is linearly calculated, and the calculation results are used as weights to obtain:

[0169]

[0170] Where R is the total number of training times;

[0171] Set the range of the objective function F(x), recorded as:

[0172]

[0173] Among them, z is the output control quantity of the control system, z n is the compensation of the neural network discriminator;

[0174] In order to minimize the objective function value, the compensation amount of the control cycle is set to t c , the output of the obstacle crossing gait controller is t o, the current control quantity t(j) of the control system is the sum of the above two. The sum value can be calculated to complete the training of the neural network model and the intelligent obstacle-crossing gait control task of the humanoid robot. The calculation formula of the control output sum is:

[0175] t(j)=t c (j)+t o (j)

[0176] Among them, j is the time required for current control.

[0177] In one embodiment, the comparison module 5 includes:

[0178] A real-time monitoring unit is used to monitor in real time the control deviation and center of mass height of the humanoid robot when it crosses an obstacle, and obtain the terrain and obstacle corresponding to the current control deviation, and obtain the obstacle corresponding to the current center of mass height;

[0179] A first comparison unit is used to compare the control deviation with a standard value of the control deviation under corresponding terrain and obstacles;

[0180] The second comparison unit is configured to form a current center of mass height fluctuation range from multiple center of mass heights of the humanoid robot when crossing an obstacle, and compare the current center of mass height fluctuation range with the center of mass height standard fluctuation range according to a set rule, wherein the set rule is:

[0181] Calculate the difference between the maximum value of the current center of mass height fluctuation range and the maximum value of the center of mass height standard fluctuation range, and compare it with the set threshold;

[0182] The highest value and the lowest value of the current center of mass height fluctuation range are compared with the highest value and the lowest value of the center of mass height standard fluctuation range.

[0183] In one embodiment, the determination module 6 includes:

[0184] a first determining unit, configured to determine that the control performance of the humanoid robot is good when the control deviation is less than a set value and the center of mass height fluctuation is normal;

[0185] a second determining unit, configured to determine that the control performance of the humanoid robot is medium when the control deviation is less than a set value and the center of mass height fluctuates abnormally;

[0186] The second determining unit is configured to determine that the control performance of the humanoid robot is poor when the control deviation is greater than or equal to a set value.

[0187] In one embodiment, in the second determining unit, the abnormal fluctuation of the center of mass height is:

[0188] Calculating the difference between the highest value of the current center of mass height fluctuation range and the highest value of the center of mass height standard fluctuation range, and when the difference is greater than or equal to a set threshold, determining that the center of mass height fluctuation is abnormal;

[0189] When the maximum value of the current center of mass height fluctuation range is higher than the maximum value of the center of mass height standard fluctuation range, and the minimum value of the current center of mass height fluctuation range is lower than the minimum value of the center of mass height standard fluctuation range, it is determined that the center of mass height fluctuation is abnormal.

[0190] The above modules and units are used to execute the corresponding steps in the above humanoid robot real-time control performance evaluation method. The specific implementation method thereof is described in the above method embodiment and will not be repeated here.

[0191] like Figure 3 As shown, the present invention also provides a computer device, which can be a server, and its internal structure can be as follows Figure 3 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store all data required for the process of the real-time control performance evaluation method of the humanoid robot. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the real-time control performance evaluation method of the humanoid robot is implemented.

[0192] Those skilled in the art will understand that Figure 3 The structure shown in is merely a block diagram of a portion of the structure related to the present application solution and does not constitute a limitation on the computer device to which the present application solution is applied.

[0193] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, any of the above-mentioned methods for evaluating the real-time control performance of a humanoid robot is implemented.

[0194] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0195] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0196] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for evaluating the real-time control performance of a humanoid robot, characterized in that: include: calculating a control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot; performing obstacle crossing control on the humanoid robot under different terrains and obstacles according to the control output to obtain a standard value of control deviation; Measuring by a tilt platform method to set multiple center of mass heights of the humanoid robot when it faces different obstacles and surmounts them, and setting the multiple center of mass heights as a standard fluctuation range of the center of mass height; A unique mapping relationship is formed between the control deviation standard value, the center of mass height standard fluctuation range and the corresponding terrain and obstacles, and a control comparison table is established; monitoring the control deviation and center of mass height of the humanoid robot in real time when crossing an obstacle, and comparing the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles; The real-time control performance of the humanoid robot is determined according to the comparison result; wherein the real-time control performance includes good, medium, and poor.

2. The method for evaluating the real-time control performance of a humanoid robot according to claim 1, wherein: The step of calculating the control output of the obstacle-crossing gait control of the humanoid robot in a set standard state according to the posture information of the humanoid robot comprises: Divide the humanoid robot into six parts: a torso, a head, a left upper limb, a right upper limb, a left lower limb, and a right lower limb, and create a part detector for the humanoid robot; Calculating correlation indices of adjacent parts of the six parts of the humanoid robot, and obtaining posture information of the humanoid robot according to the correlation indices; The posture information of the humanoid robot and the image information of the obstacle are input into a neural network model based on feedforward compensation to perform gait control, and a control output of the obstacle-crossing gait control of the humanoid robot is obtained.

3. The method for evaluating the real-time control performance of a humanoid robot according to claim 2, wherein: The step of dividing the humanoid robot into six parts, namely, a torso, a head, a left upper limb, a right upper limb, a left lower limb, and a right lower limb, and creating a humanoid robot part detector comprises: Crop the selected part from the training image, rotate it to a vertical position, and collect multiple images of the robot body as negative examples; Calculate the directional gradient histogram features of each sample image and obtain a vector Z. Vector Z plus the positive or negative label represents the corresponding sample image; The obtained vector is introduced into the AdaBoost algorithm to train a strong classifier, which is recorded as: Among them, S(·) is the sigmoid function, h i (·) is the function for calculating the length of the i-th rectangle.

4. The method for evaluating the real-time control performance of a humanoid robot according to claim 2, wherein: The step of calculating correlation indices of adjacent parts of the six parts of the humanoid robot and obtaining posture information of the humanoid robot according to the correlation indices includes: Assume that the connected edge in the connected part is g, and the connected parts are connected to each other at point k. Then the distance between the center point of edge g and point k is: Among them, x g 、y g is the horizontal and vertical coordinates of edge g; x k 、y k is the horizontal and vertical coordinates of vertex k; The correlation index is expressed as: O d =p (―1×(d―n)) / (N―n) Among them, p is the association constant, n is the minimum d value in all candidate areas, and N is the maximum d value in all candidate areas. The smaller the d value, the higher the association degree. d The larger the value of , the higher the correlation degree of the humanoid robot posture information.

5. The method for evaluating the real-time control performance of a humanoid robot according to claim 2, wherein: The step of inputting the posture information of the humanoid robot and the image information of the obstacle into a neural network model based on feedforward compensation to perform gait control, and obtaining a control output of the obstacle-crossing gait control of the humanoid robot, comprises: The image information of the obstacle is obtained by using a camera. The image information is described using a coordinate system, and the obstacle projection orientation is expressed as: Among them, u and w are the horizontal and vertical reference point orientations; Setting I (1) , O (1) are the input value and output value of the first layer of the neural network model based on feedforward compensation. In the input layer, the acceleration sensor measurement signal f of the humanoid robot and the posture information are used as input and substituted into the input end of the network controller. At the same time, the output value f is i Passed to the next layer of the network, the following corresponding relationship exists at this time, namely: O (1) =I (1) =f Inside the hidden layer, the sigmoid function is used to obtain the activation function of each node input: Among them, s i is the function weight value under the i-th node; In the output layer, the information obtained by the hidden layer is linearly calculated, and the calculation results are used as weights to obtain: Where R is the total number of training times; Set the range of the objective function F(x), recorded as: Among them, z is the output control quantity of the control system, z n is the compensation of the neural network discriminator; In order to minimize the objective function value, the compensation amount of the control cycle is set to t c , the output of the obstacle crossing gait controller is t o , the current control quantity t(j) of the control system is the sum of the above two. The sum value can be calculated to complete the training of the neural network model and the intelligent obstacle-crossing gait control task of the humanoid robot. The calculation formula of the control output sum is: t(j)=t c (j)+t o (j) Among them, j is the time required for current control.

6. The method for evaluating the real-time control performance of a humanoid robot according to claim 1, wherein: The step of monitoring the control deviation and center of mass height of the humanoid robot in real time when crossing an obstacle, and comparing the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles, includes: monitoring in real time the control deviation and center of mass height of the humanoid robot when crossing an obstacle, and obtaining the terrain and obstacle corresponding to the current control deviation, and obtaining the obstacle corresponding to the current center of mass height; Comparing the control deviation with the standard value of the control deviation under the corresponding terrain and obstacles; The multiple center-of-mass heights of the humanoid robot when crossing an obstacle are used to form a current center-of-mass height fluctuation range, and the current center-of-mass height fluctuation range is compared with the center-of-mass height standard fluctuation range according to a set rule, where the set rule is: Calculate the difference between the maximum value of the current center of mass height fluctuation range and the maximum value of the center of mass height standard fluctuation range, and compare it with the set threshold; The highest value and the lowest value of the current center of mass height fluctuation range are compared with the highest value and the lowest value of the center of mass height standard fluctuation range.

7. The method for evaluating the real-time control performance of a humanoid robot according to claim 6, wherein: The step of determining the real-time control performance of the humanoid robot according to the comparison result includes: When the control deviation is less than a set value and the center of mass height fluctuation is normal, it is determined that the control performance of the humanoid robot is good; When the control deviation is less than a set value and the center of mass height fluctuates abnormally, it is determined that the control performance of the humanoid robot is medium; When the control deviation is greater than or equal to a set value, it is determined that the control performance of the humanoid robot is poor.

8. The method for evaluating the real-time control performance of a humanoid robot according to claim 7, wherein: The abnormal fluctuation of the center of mass height is as follows: Calculating the difference between the highest value of the current center of mass height fluctuation range and the highest value of the center of mass height standard fluctuation range, and when the difference is greater than or equal to a set threshold, determining that the center of mass height fluctuation is abnormal; When the maximum value of the current center of mass height fluctuation range is higher than the maximum value of the center of mass height standard fluctuation range, and the minimum value of the current center of mass height fluctuation range is lower than the minimum value of the center of mass height standard fluctuation range, it is determined that the center of mass height fluctuation is abnormal.

9. A humanoid robot real-time control performance evaluation system, characterized in that: include: A calculation module, configured to calculate a control output of the obstacle-crossing gait control of the humanoid robot under a set standard state according to the posture information of the humanoid robot; an acquisition module, configured to perform obstacle-crossing control on the humanoid robot under different terrains and obstacles according to the control output, so as to obtain a control deviation standard value; a setting module, configured to measure and set multiple center of mass heights of the humanoid robot when facing different obstacles and overcoming them by using a tilt platform method, and to set the multiple center of mass heights as a standard fluctuation range of the center of mass height; Establishing a module for forming a unique mapping relationship between the control deviation standard value, the center of mass height standard fluctuation range and the corresponding terrain and obstacles, and establishing a control comparison table; a comparison module, configured to monitor in real time the control deviation and center of mass height of the humanoid robot when it crosses an obstacle, and compare the real-time monitored control deviation and center of mass height with corresponding standard values ​​in the control comparison table according to the terrain and obstacles; A determination module is used to determine the real-time control performance of the humanoid robot according to the comparison result; wherein the real-time control performance includes good, medium, and poor.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images