A new method for solving inverse kinematics of 7-DOF redundant manipulator was proposed.The multi-dimensional sequence processing method of closed-loop model modeling was adopted.Unscented Kalman filter algorithm was used as a recursive algorithm to estimate the real-time state of manipulator.The input vector was composed of the input sequence of forward model and the measurement output sequence of feedback model.The state space model of 7-DOF redundant manipulator inverse kinematics with closed-loop frame was established and the the computational complexity was reduced.Compared with the traditional Jacobian matrix pseudo inverse calculation method,it shows that the proposed method has higher position error accuracy and calculation efficiency.At the same time,noise was added into the trajectory of the end effector,which shows that the proposed method has better convergence performance.The experiment shows that the method has certain universality and reliability.