CPC B25J 9/1697 (2013.01) [B25J 9/1669 (2013.01)] | 6 Claims |
1. A motion capture method of a robotic arm, comprising steps of:
S1. fastening a visual sensor on a robotic arm to acquire data as a source domain, fastening an inertial sensor on a corresponding human arm to acquire data as a target domain, and establishing a state space expression of a system;
S2. setting an optimal unknown state observed joint distribution based on the state space expression and by using a total probability theory and using an observed prediction distribution of the source domain as a condition, decomposing a conditional joint observed distribution model, and solving an optimal distribution by using KL divergence; and
S3. transferring knowledge of the source domain measured by the visual sensor into the target domain measured by the inertial sensor based on a Kalman filter (KF) and the total probability theory, performing data fusion based on Kalman filtering, and predicting a state of the system at a next moment to implement motion capture of the robotic arm.
|