4 results about "Joint moment" patented technology

The invention discloses a gait muscle force and activation real-time prediction method based on a physical constraint model, and belongs to the technical field of sports biomechanics and artificial intelligence. The method comprises the following steps: 1, inputting only hip, knee and ankle joint angles as input data; 2, a model backbone is a Transform; 3, multi-task loss constraint model training; 4, adjusting model parameters; and 5, calculating muscle activation. According to the invention, force and activation of multiple muscles can be output in real time only by inputting a joint angle sequence; the joint moment constraint is added in the loss, so that the prediction meets the biomechanical law, the precision, robustness and interpretability are considered, the traditional static optimization or dynamic optimization method can be replaced, and the process can be carried out in multiple scenes.

The invention relates to the technical field of robot control, and discloses an underactuated walking-aided exoskeleton robot control method, system and terminal for a body rising scene, and the method comprises the steps: constructing an underactuated walking-aided exoskeleton robot, simplifying the underactuated walking-aided exoskeleton robot to obtain a four-bar system, and connecting the four-bar system with the underactuated walking-aided exoskeleton robot; a trunk hip joint angle, a knee joint angle and an ankle joint angle in a four-bar system are obtained through an inertial measurement unit and an inverse kinematics technology, and a moment of each joint of a lower limb is iteratively solved through a Newton-Euler method to obtain a knee joint moment and a hip joint moment; calculating the required motor torque according to the knee joint torque and the hip joint torque, and performing sole contact detection on the motor torque to obtain a command torque; according to the current coefficient of the motor, the command torque is converted into command current, and the command current is input into a controller of the motor to control the under-actuated walking-aid exoskeleton robot. According to the method, assistance of the multi-joint exoskeleton can be more robust and stable.

The application relates to the technical field of robot control, and discloses an underactuated walking exoskeleton robot control method, system and terminal for a body ascending scene, the method comprising the following steps: constructing an underactuated walking exoskeleton robot, simplifying the underactuated walking exoskeleton robot to obtain a four-bar system, acquiring a trunk hip joint angle, a knee joint angle and an ankle joint angle in the four-bar system through an inertial measurement unit and inverse kinematics technology, iteratively solving the moments of each joint of a lower limb through a Newton-Euler method to obtain a knee joint moment and a hip joint moment, calculating required motor moments according to the knee joint moment and the hip joint moment, performing foot contact detection on the motor moments to obtain a command moment; and converting the command moment into a command current according to a current coefficient of the motor, and inputting the command current into a controller of the motor to control the underactuated walking exoskeleton robot. The application makes the assistance of a multi-joint exoskeleton more robust and stable.

This invention discloses a method for estimating lower limb joint moments based on deep learning using physical information, belonging to the field of lower limb joint moment estimation technology. The method includes: acquiring data from three six-axis inertial measurement units (IMUs) set on the lower limb during movement and corresponding reference joint moment data based on an open-source lower limb biomechanical dataset; constructing a temporal input after preprocessing; building a TCN-BiLSTM model combining a temporal convolutional network and a bidirectional long short-term memory network; establishing a loss function containing physical information constraints based on prior knowledge of lower limb dynamics; training the model using the loss function; and simultaneously estimating the joint moments of the hip, knee, and ankle joints in the sagittal plane using the trained model. Compared with existing technologies, this invention can achieve high-precision estimation of lower limb multi-joint moments under various movement modes with a small amount of inertial sensor data input, improving the model's physical consistency, generalization ability, and ease of application.