Finger capture device and method for humanoid robot reinforcement learning

Abstract

The present application relates to the field of robotics technology, and discloses a finger capture device and method for reinforcement learning of a humanoid robot, comprising: a finger module, arranged on each finger of the humanoid robot, comprising an IMU sensor, an LED light, a tactile sensor and a processing module; the IMU sensor is used to collect motion state data of the finger; the LED light is used to perform ID identification on each finger through its flashing sequence; the processing module is used to process data obtained by the IMU sensor and control the LED light; the tactile sensor is used to sense the contact state between the finger and an external object; a camera is used to obtain two-dimensional image data of the humanoid robot's hand position in multiple directions; an analysis center is used to identify the spatial position coordinates of each finger based on the two-dimensional image data in multiple directions, and to judge the hand posture based on the spatial position coordinates of each finger, the contact state with the external object and the motion state data, thereby improving the anti-interference and accuracy of optical motion capture.

Description

Technical Field

[0001] The present application relates to the technical field of robot finger motion capture, and in particular to a finger capture device and method for reinforcement learning of a humanoid robot. Background Art

[0002] With the development of robotics, the application of humanoid robots in reinforcement learning is becoming increasingly widespread. During the learning process of humanoid robots, precise finger tracking and control are key to achieving various complex tasks. By acquiring finger posture, force state, and motion data, the humanoid robot can be better controlled and managed.

[0003] Existing finger capture methods for humanoid robots include inertial motion capture, optical motion capture, and tactile and force feedback. Among them, the inertial motion capture system uses micro sensors to record finger motion data, combines AI technology to train robot motion, and realizes the motion simulation process. It also uses integrated fingertip tracking sensors to capture finger bending, stretching and other movements in real time, and then combines tactile feedback and machine learning algorithms to optimize and improve the realism of the robot's hand movements. Optical motion capture uses multiple cameras to track feature points to realize the collection of whole body and hand movements.

[0004] Among the existing finger capture methods, the single acquisition method has problems such as low accuracy and poor stability. Among them, optical motion capture is easily affected by external interference, resulting in large errors in the finger joint tracking process. Therefore, how to improve the anti-interference ability and accuracy of optical motion capture is the fundamental problem to be solved by the present invention. Summary of the Invention

[0005] In order to improve the anti-interference ability and accuracy of optical motion capture, the present application provides a finger capture device and method for reinforcement learning of humanoid robots.

[0006] In a first aspect, the present application provides a finger capture device for reinforcement learning of a humanoid robot, which adopts the following technical solution:

[0007] A finger capture device for reinforcement learning of a humanoid robot, comprising:

[0008] The finger module is installed on each finger of the humanoid robot and includes an IMU sensor, an LED light, a tactile sensor, and a processing module. The IMU sensor is used to collect the motion state data of the finger. The LED light is used to identify each finger by its flashing sequence. The processing module is used to process the data obtained by the IMU sensor and control the LED light. The tactile sensor is used to sense the contact state between the finger and the external object.

[0009] A microcontroller, which receives and processes IMU data from the finger module and transmits it to the analysis center;

[0010] The camera is provided in several groups and is used to obtain two-dimensional image data of multiple orientations of the humanoid robot's hand positions;

[0011] The analysis center is used to obtain the spatial position coordinates of each finger based on two-dimensional image data in multiple directions, and to judge the hand posture based on the spatial position coordinates of each finger, the contact status with external objects and the motion status data.

[0012] By adopting the above technical solution, the finger where the LED light is located is determined by the flashing sequence of the LED light, which avoids the influence of external interference factors on the judgment of the position point, thereby improving the accuracy of the acquisition of the LED and other position points, and using the spatial position coordinates of each finger, the contact status with external objects and the motion status data to make a comprehensive judgment on the hand posture, avoiding the problems of large error rate and inaccurate acquisition results in a single acquisition method.

[0013] Optionally, the process of controlling the flashing sequence of several groups of LED lights to perform ID recognition on each finger includes:

[0014] Unique flashing sequences are pre-set for different fingers, and the ID of each finger is parsed through continuous two-dimensional image data.

[0015] By adopting the above technical solution, each finger can have a unique flashing sequence, and the serial number of the finger can be identified through the flashing sequence, thereby ensuring the anti-interference and accuracy of the finger spatial coordinate position acquisition.

[0016] Optionally, the process of the analysis center judging the hand posture includes:

[0017] The position of the LED light is used as the feature point, the feature point in the two-dimensional image data is extracted, and the three-dimensional coordinates of the feature point are obtained based on the parallax calculation of multiple azimuth feature points;

[0018] Perform stereo matching screening based on the 3D coordinates of the feature points and the preset posture set in the database to obtain a 3D model of the hand;

[0019] The hand posture is judged based on the three-dimensional model of the hand and the contact status and motion status data of each finger with external objects.

[0020] By adopting the above technical solution, the hand posture is judged based on the three-dimensional model of the hand and the contact status and motion status data of each finger with external objects, thereby improving the anti-interference ability of finger status acquisition and obtaining more accurate finger position and posture information.

[0021] Optionally, each finger module is provided with two groups of LED lights, and the two groups of LED lights are provided in different finger segments. The stereo matching screening process includes:

[0022] Obtain the spatial distances between the LED lights of different finger segments on each finger to form a first distance sequence [x1, x2, ..., xm]; perform a first comparison between the first distance sequence [x1, x2, ..., xm] and the first distance sequence [xi1, xi2, ..., xim] corresponding to each preset posture in the preset posture set, and then Sort each preset posture in ascending order, extract the preset posture according to the preset screening rule to obtain the first screening set; where m is the number of fingers, k is a positive integer and k∈[1,m], is the length influence coefficient of the k-th finger;

[0023] Get the midpoint of the line connecting the LED lights of different finger segments on each finger, and obtain the distance between the midpoints of the line connecting adjacent fingers according to the order of fingers to form a second distance sequence [y1, y2, ..., y(m-1)]; perform a second comparison on the second distance sequence [y1, y2, ..., y(m-1)] corresponding to each preset posture in the first screening set [yi1, yi2, ..., yi(m-1)], and then calculate the distance between the midpoints of the line connecting adjacent fingers according to the order of fingers. The values ​​are sorted from small to large, and the top Q are selected for the first vector comparison process. The corresponding preset posture is obtained as the stereo matching screening result according to the first vector comparison process, and the corresponding hand 3D model is obtained. is the proportional adjustment coefficient, j∈[1,m-1].

[0024] By adopting the above technical solution, the amount of calculation in the screening process is reduced and the screening efficiency is improved while ensuring the accuracy of the screening results.

[0025] Optionally, the preset screening rules include:

[0026] judge The corresponding number g, if the number g exceeds the preset value gt, then the top gt names in the sorting are selected as the first screening set, otherwise, the top g names in the sorting are selected as the first screening set, where Xt is the preset threshold.

[0027] By adopting the above technical solution, the preset postures in the first screening set are obtained through preset screening rules, which can ensure the accuracy and comprehensiveness of the obtained results.

[0028] Optionally, the first vector comparison process includes:

[0029] The midpoint of the line connecting the LED lights of different finger segments on the first finger and the midpoint of the line connecting the LED lights of different finger segments on other fingers form the first vector sequence , the first vector sequence Compare with the first vector sequence of the Q preset postures before sorting, and select The preset posture corresponding to the minimum value is taken as the stereo matching screening result, where s∈[1,m-2], is the sth vector in the first vector sequence The angle of the first vector sequence corresponding to the preset posture, is the bias error coefficient.

[0030] By adopting the above technical solution, the corresponding preset posture can be accurately matched.

[0031] Optionally, each finger module is provided with a set of LED lights, and the stereo matching screening process includes:

[0032] Obtain the distances between adjacent LED lights in the order of fingers to form a third distance sequence [z1, z2, ..., z(m-1)]; perform a third comparison on the third distance sequence [z1, z2, ..., z(m-1)] with the third distance sequence [zi1, zi2, ..., zi(m-1)] corresponding to each preset posture in the first screening set, and then The values ​​are sorted from small to large, and the top Q are selected for the second vector comparison process. The corresponding preset posture is obtained according to the second vector comparison process as the stereo matching screening result, and the corresponding three-dimensional hand model is obtained.

[0033] By adopting the above technical solution, the corresponding preset posture can be screened more quickly, thereby improving the efficiency of the screening process.

[0034] Optionally, the second vector comparison process includes:

[0035] The coordinates of the LED light on the first finger and the coordinates of the LED lights on other fingers form a second vector sequence , the second vector sequence Compare with the first vector sequence of the Q preset postures before sorting, and select The preset pose corresponding to the minimum value is taken as the stereo matching screening result, where p∈[1,m-2], is the pth vector in the second vector sequence The angle of the first vector sequence corresponding to the preset posture, is the bias error coefficient.

[0036] By adopting the above technical solution, the corresponding preset posture can be accurately matched.

[0037] In a second aspect, the present application provides a finger capture method for reinforcement learning of a humanoid robot, which adopts the following technical solution:

[0038] A finger capture method for reinforcement learning of a humanoid robot, the method using any of the above-mentioned finger capture devices for reinforcement learning of a humanoid robot, comprising:

[0039] The IMU sensor is used to collect the motion status data of the finger; the LED light flashing sequence is used to identify each finger; the processing module is used to process the data obtained by the IMU sensor and control the LED light; the tactile sensor is used to sense the contact status between the finger and the external object;

[0040] The IMU data from the finger module is received and processed by the microcontroller and transmitted to the analysis center;

[0041] Acquire two-dimensional image data of multiple positions of the humanoid robot's hands through several groups of cameras;

[0042] The analysis center obtains the spatial position coordinates of each finger based on the two-dimensional image data in multiple directions, and judges the hand posture based on the spatial position coordinates of each finger, the contact status with external objects and the motion status data.

[0043] In summary, this application includes at least one of the following beneficial technical effects:

[0044] The present invention can determine the finger where the LED light is located through the flashing sequence of the LED light, avoiding the influence of external interference factors on the judgment of the position point, thereby improving the accuracy of the collection of position points such as the LED, and using the spatial position coordinates of each finger, the contact state with the external object and the motion state data to comprehensively judge the hand posture of the humanoid robot, avoiding the problems of large error rate and inaccurate collection results in a single collection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a logic block diagram of a finger capture device for reinforcement learning of a humanoid robot according to the present invention.

[0046] Figure 2 Each finger module is equipped with two sets of LED lights to show the finger capture effect.

[0047] Figure 3 Each finger module is equipped with a set of LED lights to show the finger capture effect.

[0048] Figure 4 It is a flow chart of the finger capture method for reinforcement learning of a humanoid robot according to the present invention. DETAILED DESCRIPTION

[0049] Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings.

[0050] Throughout this specification, reference to the terms "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0051] The present application embodiment discloses a finger capture device for humanoid robot reinforcement learning, referring to Figure 1 , including a finger module, a microcontroller, several groups of cameras and an analysis center, wherein the finger module is set on each finger of the humanoid robot, including an IMU sensor, an LED light, a tactile sensor and a processing module; the IMU sensor is used to collect the motion state data of the finger, and the motion state data includes inertial data such as acceleration and angular velocity of the finger. The LED light is used to ID each finger through its flashing sequence. The flashing sequence of the LED light is pre-set according to the serial number of different fingers. The color and brightness of the LED light can be selected according to actual needs. The finger where the LED light is located can be determined through the flashing sequence of the LED light, which avoids the influence of external interference factors on the judgment of the position point, thereby improving the accuracy of the acquisition of the LED and other position points. At the same time, the processing module is used to process the data obtained by the IMU sensor and control the LED light; the tactile sensor is used to sense the hand The contact state of the finger with the external object, the tactile sensor is installed at the end of each finger, which can sense the contact force between the finger and the external object. The tactile sensor can be selected from piezoelectric, capacitive or resistive types, and the appropriate sensor is selected according to the specific application scenario; in addition, the microcontroller is used to receive and process the IMU data from the finger module and transmit it to the analysis center; there are several groups of cameras, which can capture the two-dimensional image information of the position of the humanoid robot's hand in different directions. After that, the analysis center uses the two-dimensional image data in multiple directions to obtain the spatial position coordinates of each finger, and then judge the hand posture according to the spatial position coordinates of each finger, the contact state with the external object and the motion state data. The above data is used to make a comprehensive judgment on the hand posture of the humanoid robot, avoiding the problems of large error rate and inaccurate collection results in a single collection method.

[0052] In one embodiment, a comprehensive judgment process is provided. The three-dimensional coordinate data is filtered to remove noise points, and the obtained three-dimensional coordinate data is screened and matched with preset information in a database to obtain a corresponding three-dimensional model of the finger. The IMU data is calibrated and filtered to eliminate errors and improve the accuracy of posture calculation. A threshold judgment is performed on the tactile sensor data to determine whether effective contact has occurred. The pre-processed three-dimensional coordinate data is fused with the IMU data by optical inertia, and a Kalman filter algorithm is used to perform state estimation with the three-dimensional model and the state transition model provided by the IMU data as the system model. Through continuous iterative updates, more accurate finger position and posture information is obtained.

[0053] In one embodiment, a process is provided for controlling a plurality of LED light flashing sequences to perform ID identification on each finger, including pre-setting corresponding unique flashing sequences for different fingers, and parsing the ID of each finger through continuous two-dimensional image data. Through this process, each finger can have a unique flashing sequence, and the serial number of the finger can be identified through the flashing sequence, thereby ensuring the anti-interference and accuracy of the acquisition of the spatial coordinate position of the finger.

[0054] In one embodiment, a process for the analysis center to judge the hand posture is given, including: taking the position of the LED light as the feature point, extracting the feature point in the two-dimensional image data, and obtaining the three-dimensional coordinates of the feature point based on the parallax calculation of multiple orientation feature points. The implementation of this process is based on the three-dimensional reconstruction technology of multiple cameras, which will not be further described here. Stereo matching and screening are performed based on the three-dimensional coordinates of the feature point and the preset posture set in the database to obtain a three-dimensional model of the hand; the hand posture is judged based on the three-dimensional model of the hand and the contact status and motion status data of each finger with external objects, thereby improving the anti-interference ability of the finger status acquisition and obtaining more accurate finger position and posture information.

[0055] In one embodiment, see the attached Figure 2 , each finger module is provided with two groups of LED lights, and the two groups of LED lights are provided in different finger segments. The stereo matching screening process under this setting includes: obtaining the spatial distances of the LED lights of different finger segments on each finger to form a first distance sequence [x1, x2, ..., xm]; performing a first comparison on the first distance sequence [x1, x2, ..., xm] with the first distance sequence [xi1, xi2, ..., xim] corresponding to each preset posture in the preset posture set; according to Sort each preset posture in order from small to large, where is the length influence coefficient of the kth finger, which is related to the finger length of the humanoid robot. When the finger length is longer, the deviation generated will be larger, and the deviation of the shorter finger will have a greater impact on the overall result than that of the longer finger. Therefore, the length influence coefficient is negatively correlated with the finger length of the humanoid robot. When the finger lengths of all humanoid robots are the same, the length influence coefficient is also the same. The preset postures are extracted according to the preset screening rules to obtain the first screening set; wherein m is the number of fingers, k is a positive integer and k∈[1, m]. Through the first comparison process, the preliminary screening process of the preset postures in the database can be realized. Then, the midpoint of the line connecting the LED lights of different finger segments on each finger is obtained, and the distance between the midpoints of the lines connecting adjacent fingers is obtained according to the order of the fingers to form a second distance sequence [y1, y2, …, y(m-1)]; the second distance sequence [y1, y2, …, y(m-1)] is compared with the second distance sequence [yi1, yi2, …, yi(m-1)] corresponding to each preset posture in the first screening set for the second time. The values ​​are sorted from small to large, where is the proportional adjustment coefficient, which is obtained according to the fitting setting of empirical data. In addition, j∈[1, m-1], the top Q ranked ones are selected to perform the first vector comparison process, and then the secondary screening process is completed. According to the first vector comparison process, the corresponding preset posture is obtained as the stereo matching screening result, and the corresponding three-dimensional hand model is obtained. Through the above multi-level screening process, the accuracy of the screening results can be guaranteed while reducing the amount of calculation in the screening process and improving the screening efficiency.

[0056] In one embodiment, a preset screening rule is provided, including: judging The corresponding number g, Xt is a preset threshold, which is set according to empirical data. The number g is compared with the preset value gt. The preset value gt is set according to empirical data. If the number g exceeds the preset value gt, it means that there are more matching items. Therefore, the top gt names in the sorting are selected as the first screening set to ensure the number of matching preset postures in the first screening set. Otherwise, the top g names in the sorting are selected as the first screening set to ensure the validity and matching of the matching preset postures in the first screening set. The preset postures in the first screening set are obtained through the preset screening rules, which can ensure the accuracy and comprehensiveness of the obtained results.

[0057] In one embodiment, a first vector comparison process is provided, including: forming a first vector sequence by combining the midpoints of the line connecting the LED lights of different finger segments on the first finger with the midpoints of the line connecting the LED lights of different finger segments on other fingers. , the first vector sequence Compare with the first vector sequence of the Q preset postures before sorting, and select The preset posture corresponding to the minimum value is taken as the stereo matching screening result, where s∈[1,m-2], is the sth vector in the first vector sequence The angle of the first vector sequence corresponding to the preset posture, is the bias error coefficient, which is set according to the distance between the other fingers and the first finger. The farther the distance between the two fingers, the greater the angular deviation will be, and the closer the distance, the greater the impact of the angular deviation on the overall result compared to the shorter finger. Therefore, the bias error coefficient is negatively correlated with the finger distance of the humanoid robot. Through the above-mentioned first vector comparison process, the corresponding preset posture can be accurately matched.

[0058] In one embodiment, see the attached Figure 3 , each finger module is provided with a group of LED lights, and the stereo matching screening process includes: obtaining the distances of adjacent LED lights in the order of fingers to form a third distance sequence [z1, z2, ..., z(m-1)]; performing a third comparison on the third distance sequence [z1, z2, ..., z(m-1)] with the third distance sequence [zi1, zi2, ..., zi(m-1)] corresponding to each preset posture in the first screening set, and performing a third comparison according to the third distance sequence [zi1, zi2, ..., zi(m-1)]. The values ​​are sorted from small to large, and the top Q are selected for the second vector comparison process. The corresponding preset posture is obtained according to the second vector comparison process as the stereo matching screening result, and the corresponding three-dimensional hand model is obtained. When each finger module is provided with a group of LED lights, the corresponding preset posture can be screened and obtained more quickly through the third comparison process and the second vector comparison process, thereby improving the efficiency of the screening process.

[0059] In one embodiment, a second vector comparison process is provided, including: forming a second vector sequence by combining the coordinates of the LED light on the first finger with the coordinates of the LED lights on other fingers. , the second vector sequence Compare with the first vector sequence of the Q preset postures before sorting, and select The preset pose corresponding to the minimum value is taken as the stereo matching screening result, where p∈[1,m-2], is the pth vector in the second vector sequence The angle of the first vector sequence corresponding to the preset posture, is the bias error coefficient. Through the above second vector comparison process, the corresponding preset posture can be accurately matched.

[0060] The present application also discloses a finger capture method for humanoid robot reinforcement learning, referring to Figure 4, including: using an IMU sensor to collect finger motion state data; using an LED light flashing sequence to identify each finger; using a processing module to process the data obtained by the IMU sensor and control the LED light; using a tactile sensor to sense the contact state of the finger with an external object; using a microcontroller to receive and process the IMU data from the finger module and transmit it to an analysis center; using a plurality of cameras to obtain two-dimensional image data of multiple orientations of the humanoid robot's hand position; using the analysis center to obtain the spatial position coordinates of each finger based on the two-dimensional image data in multiple orientations, and judging the hand posture based on the spatial position coordinates of each finger, the contact state with the external object, and the motion state data. This embodiment can determine the finger where the LED light is located through the LED light flashing sequence, avoiding the influence of external interference factors on the judgment of the position point, thereby improving the accuracy of the acquisition of the position point such as the LED, and using the spatial position coordinates of each finger, the contact state with the external object, and the motion state data to comprehensively judge the hand posture of the humanoid robot, avoiding the problems of a large error rate and inaccurate acquisition results in a single acquisition method.

[0061] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A finger capture device for reinforcement learning of a humanoid robot, characterized in that: include: The finger module is installed on each finger of the humanoid robot and includes an IMU sensor, an LED light, a tactile sensor, and a processing module. The IMU sensor is used to collect the motion state data of the finger. The LED light is used to identify each finger by its flashing sequence. The processing module is used to process the data obtained by the IMU sensor and control the LED light. The tactile sensor is used to sense the contact state between the finger and the external object. A microcontroller, which receives and processes IMU data from the finger module and transmits it to the analysis center; The camera is provided in several groups and is used to obtain two-dimensional image data of multiple orientations of the humanoid robot's hand positions; The analysis center is used to obtain the spatial position coordinates of each finger based on the two-dimensional image data in multiple directions, and to judge the hand posture based on the spatial position coordinates of each finger, the contact state with the external object, and the motion state data; The process of controlling the flashing sequence of several LED lights to identify each finger ID includes: Unique flashing sequences are pre-set for different fingers, and the ID of each finger is parsed through continuous two-dimensional image data; The process of the analysis center judging the hand posture includes: The position of the LED light is used as the feature point, the feature point in the two-dimensional image data is extracted, and the three-dimensional coordinates of the feature point are obtained based on the parallax calculation of multiple azimuth feature points; Perform stereo matching screening based on the 3D coordinates of the feature points and the preset posture set in the database to obtain a 3D model of the hand; The hand posture is judged based on the three-dimensional model of the hand and the contact status and motion status data of each finger with external objects.

2. The finger capture device for reinforcement learning of a humanoid robot according to claim 1, characterized in that: Each finger module is provided with two sets of LED lights, and the two sets of LED lights are set in different finger segments. The stereo matching screening process includes: Obtain the spatial distances between the LED lights of different finger segments on each finger to form a first distance sequence [x1, x2, ..., xm]; perform a first comparison between the first distance sequence [x1, x2, ..., xm] and the first distance sequence [xi1, xi2, ..., xim] corresponding to each preset posture in the preset posture set, and then Sort each preset posture in ascending order, extract the preset posture according to the preset screening rule to obtain the first screening set; where m is the number of fingers, k is a positive integer and k∈[1,m], is the length influence coefficient of the k-th finger; Get the midpoint of the line connecting the LED lights of different finger segments on each finger, and obtain the distance between the midpoints of the line connecting adjacent fingers according to the order of fingers to form a second distance sequence [y1, y2, ..., y(m-1)]; perform a second comparison on the second distance sequence [y1, y2, ..., y(m-1)] corresponding to each preset posture in the first screening set [yi1, yi2, ..., yi(m-1)], and then calculate the distance between the midpoints of the line connecting adjacent fingers according to the order of fingers. The values ​​are sorted from small to large, and the top Q are selected for the first vector comparison process. The corresponding preset posture is obtained as the stereo matching screening result according to the first vector comparison process, and the corresponding hand 3D model is obtained. is the proportional adjustment coefficient, j∈[1,m-1].

3. The finger capture device for reinforcement learning of a humanoid robot according to claim 2, characterized in that: The preset screening rules include: judge The corresponding number g, if the number g exceeds the preset value gt, then the top gt names in the sorting are selected as the first screening set, otherwise, the top g names in the sorting are selected as the first screening set, where Xt is the preset threshold.

4. The finger capture device for reinforcement learning of a humanoid robot according to claim 2, characterized in that: The first vector comparison process includes: The midpoint of the line connecting the LED lights of different finger segments on the first finger and the midpoint of the line connecting the LED lights of different finger segments on other fingers form the first vector sequence , the first vector sequence Compare with the first vector sequence of the Q preset postures before sorting, and select The preset posture corresponding to the minimum value is taken as the stereo matching screening result, where s∈[1,m-2], is the sth vector in the first vector sequence The angle of the first vector sequence corresponding to the preset posture, is the bias error coefficient.

5. The finger capture device for reinforcement learning of a humanoid robot according to claim 1, characterized in that: Each finger module is provided with a set of LED lights. The stereo matching screening process includes: Obtain the distances between adjacent LED lights in the order of fingers to form a third distance sequence [z1, z2, ..., z(m-1)]; perform a third comparison on the third distance sequence [z1, z2, ..., z(m-1)] with the third distance sequence [zi1, zi2, ..., zi(m-1)] corresponding to each preset posture in the first screening set, and then The values ​​are sorted from small to large, and the top Q are selected for the second vector comparison process. The corresponding preset posture is obtained according to the second vector comparison process as the stereo matching screening result, and the corresponding three-dimensional hand model is obtained.

6. The finger capture device for reinforcement learning of a humanoid robot according to claim 5, characterized in that: The second vector comparison process includes: The coordinates of the LED light on the first finger and the coordinates of the LED lights on other fingers form a second vector sequence , the second vector sequence Compare with the first vector sequence of the Q preset postures before sorting, and select The preset pose corresponding to the minimum value is taken as the stereo matching screening result, where p∈[1,m-2], is the pth vector in the second vector sequence The angle of the first vector sequence corresponding to the preset posture, is the bias error coefficient.

7. A finger capture method for reinforcement learning of a humanoid robot, characterized in that: The method uses a finger capture device for reinforcement learning of a humanoid robot according to any one of claims 1 to 6, comprising: The IMU sensor is used to collect the motion status data of the finger; the LED light flashing sequence is used to identify each finger; the processing module is used to process the data obtained by the IMU sensor and control the LED light; the tactile sensor is used to sense the contact status between the finger and the external object; The IMU data from the finger module is received and processed by the microcontroller and transmitted to the analysis center; Acquire two-dimensional image data of multiple positions of the humanoid robot's hands through several groups of cameras; The analysis center obtains the spatial position coordinates of each finger based on the two-dimensional image data in multiple directions, and judges the hand posture based on the spatial position coordinates of each finger, the contact status with external objects and the motion status data.

Images