METHOD FOR OPTIMIZING HUMAN BODY POSTURE RECOGNITION MODEL, DEVICE AND COMPUTER-READABLE STORAGE MEDIUM

    公开(公告)号:US20230334893A1

    公开(公告)日:2023-10-19

    申请号:US18214467

    申请日:2023-06-26

    CPC classification number: G06V40/103 G06V10/806

    Abstract: A method includes: obtaining heat maps including a predetermined number of key points of a human body; performing depth separable convolution on a feature map corresponding to one of the heat maps corresponding to each of the key points and a convolution kernel of a corresponding channel of the human body posture recognition model to determine a key point feature map corresponding to each channel of the human body posture recognition model; performing local feature fusion processing and/or global feature fusion processing on the key point feature map corresponding to each channel to obtain fusion posture feature maps; determining a linear relationship between the channels of the human body posture recognition model based on the fusion posture feature maps; and updating weight coefficients of the corresponding channels of the human body posture recognition model by using the linear relationship between the channels of the human body posture recognition model.

    FEEDFORWARD CONTROL METHOD FOR FLOATING BASE DYNAMICS, COMPUTER-READABLE STORAGE MEDIUM AND ROBOT

    公开(公告)号:US20220019196A1

    公开(公告)日:2022-01-20

    申请号:US17088596

    申请日:2020-11-04

    Abstract: A feedforward control method comprising steps of: acquiring kinematic parameters of each joint of a robot based on inverse kinematics according to a pre-planned robot motion trajectory, and setting a center of a body of the robot as a floating base; determining a six-dimensional acceleration of a center of mass of each joint of the robot in a base coordinate system using a forward kinematics algorithm, based on the kinematic parameters of each joint of the robot, and converting the six-dimensional acceleration of the center of mass of each joint of the robot in the base coordinate system to a six-dimensional acceleration in a world coordinate system; and calculating a torque required by a motor of each joint of the robot using an inverse dynamic algorithm, and controlling the motors of corresponding joints of the robot.

    TASK HIERARCHICAL CONTROL METHOD, AND ROBOT AND COMPUTER READABLE STORAGE MEDIUM USING THE SAME

    公开(公告)号:US20220009093A1

    公开(公告)日:2022-01-13

    申请号:US17192906

    申请日:2021-03-05

    Abstract: A task hierarchical control method as well as a robot and a storage medium using the same are provided. The method includes: obtaining a task instruction for a robot, where the task instruction is for determining a target task card including an amount of selection matrices for dividing a target task into the amount of hierarchical subtasks and a controller name for executing each of the hierarchical subtasks; obtaining a null space projection matrix of each of the hierarchical subtasks based on the corresponding selection matrix; generating control finks of the amount according to the corresponding controller of each of the hierarchical subtasks and the corresponding null space projection matrix; calculating a control torque of each of the control links and obtaining a hierarchical control output quantity by adding ail the control torques; and controlling the robot to perform the target task using the hierarchical control output quantity.

    JOINT CONTROL METHOD, COMPUTER-READABLE STORAGE MEDIUM AND MULTI-LEGGED ROBOT

    公开(公告)号:US20210387332A1

    公开(公告)日:2021-12-16

    申请号:US17345005

    申请日:2021-06-11

    Abstract: A method for a multi-legged robot having a body and a number of legs, includes: obtaining a current pose of the body, forces applied to the body, and joint angles of each of supporting legs of the legs; creating a mapping matrix from the forces applied to the body to desired support forces applied to soles of the supporting legs; obtaining priority targets by prioritizing the forces acting in different directions, determining a weight matrix for each priority target, and creating an optimization model of the support forces for each priority target based on the mapping matrix and the weight matrices; solving the optimization model of each of the priority targets to obtain the desired support forces corresponding to each of the priority targets; and calculating joint torques of the supporting legs for joint control, based on the solved desired support forces and Jacobian matrices corresponding to the supporting legs.

    Multi-channel feature map fusion
    5.
    发明授权

    公开(公告)号:US12094084B2

    公开(公告)日:2024-09-17

    申请号:US17388043

    申请日:2021-07-29

    Abstract: Image processing methods are provided. One of the method includes: obtaining a to-be-processed multi-channel feature maps; obtaining multi-channel first output feature maps and multi-channel second output feature maps by processing the multi-channel feature maps through a parallel pointwise convolution and non-pointwise operation, where the non-pointwise convolution is for descripting a spatial feature of each channel and an information exchange between the feature maps; and fusing the multi-channel first output feature maps and the multi-channel second output feature maps to obtain a multi-channel third output feature map.

    Total centroid state estimation method, humanoid robot and computer readable storage medium using the same

    公开(公告)号:US11872701B2

    公开(公告)日:2024-01-16

    申请号:US17485412

    申请日:2021-09-25

    CPC classification number: B25J9/1664 B25J9/1607 G05D2201/0217

    Abstract: A total centroid state estimation method as well as a humanoid robot and a computer readable storage medium using the same are provided. The method includes: obtaining a motion state of each real joint of the humanoid robot and a motion state of its floating base, where the floating base is equivalent to a plurality of sequent-connected virtual joints; calculating a joint position, a centroid position, and a rotation matrix of each link in the world coordinate system in sequence using the chain rule of homogeneous multiplication according to the position of the joint corresponding to the link to solve a Jacobian matrix of the centroid of the link; solving a total centroid Jacobian matrix based on the Jacobian matrix of the centroid of each link and the total mass; and calculating the total centroid velocity based on the total centroid Jacobian matrix and other parameters.

    METHOD FOR CONTROLLING ROBOT, ROBOT AND COMPUTER-READABLE STORAGE MEDIUM

    公开(公告)号:US20230373089A1

    公开(公告)日:2023-11-23

    申请号:US18230620

    申请日:2023-08-05

    CPC classification number: B25J9/1661

    Abstract: A method for controlling a robot includes: obtaining current motion state information of the robot and desired motion trajectory information corresponding to a target task; determining task execution coefficient matrices corresponding to the robot performing the target task according to the desired motion trajectory information and the motion state information; constructing matching dynamic constraints for task-driven parameters of the robot according to the desired motion trajectory information and the motion state information; constructing matching parameter distribution constraints for the task-driven parameters according to the motion state information and body action safety constraints corresponding to the target task; solving a pre-stored task execution loss function by using the task execution coefficient matrices to obtain the target-driven parameters satisfying the dynamic constraints and the parameter distribution constraints; and controlling operation state of each joint end effector of the robot according to the target-driven parameters.

    Task hierarchical control method, and robot and computer readable storage medium using the same

    公开(公告)号:US11602844B2

    公开(公告)日:2023-03-14

    申请号:US17192906

    申请日:2021-03-05

    Abstract: A task hierarchical control method as well as a robot and a storage medium using the same are provided. The method includes: obtaining a task instruction for a robot, where the task instruction is for determining a target task card including an amount of selection matrices for dividing a target task into the amount of hierarchical subtasks and a controller name for executing each of the hierarchical subtasks; obtaining a null space projection matrix of each of the hierarchical subtasks based on the corresponding selection matrix; generating control finks of the amount according to the corresponding controller of each of the hierarchical subtasks and the corresponding null space projection matrix; calculating a control torque of each of the control links and obtaining a hierarchical control output quantity by adding ail the control torques; and controlling the robot to perform the target task using the hierarchical control output quantity.

    Feedforward control method for floating base dynamics, computer-readable storage medium and robot

    公开(公告)号:US11579591B2

    公开(公告)日:2023-02-14

    申请号:US17088596

    申请日:2020-11-04

    Abstract: A feedforward control method comprising steps of: acquiring kinematic parameters of each joint of a robot based on inverse kinematics according to a pre-planned robot motion trajectory, and setting a center of a body of the robot as a floating base; determining a six-dimensional acceleration of a center of mass of each joint of the robot in a base coordinate system using a forward kinematics algorithm, based on the kinematic parameters of each joint of the robot, and converting the six-dimensional acceleration of the center of mass of each joint of the robot in the base coordinate system to a six-dimensional acceleration in a world coordinate system; and calculating a torque required by a motor of each joint of the robot using an inverse dynamic algorithm, and controlling the motors of corresponding joints of the robot.

    Scene data obtaining method and model training method, apparatus and computer readable storage medium using the same

    公开(公告)号:US11461958B2

    公开(公告)日:2022-10-04

    申请号:US17216719

    申请日:2021-03-30

    Abstract: A scene data obtaining method as well as a model training method and a computer readable storage medium using the same are provided. The method includes: building a virtual simulation scene corresponding to an actual scene, where the virtual simulation scene is three-dimensional; determining a view frustum corresponding to preset view angles in the virtual simulation scene; collecting one or more two-dimensional images in the virtual simulation scene and ground truth object data associated with the one or more two-dimensional images using the view frustum corresponding to the preset view angles; and using all the two-dimensional images and the ground truth object data associated with the one or more two-dimensional images as scene data corresponding to the actual scene. In this manner, the data collection does not require manual annotation, and the obtained data can be used for training deep learning-based perceptual models.

Patent Agency Ranking