In order to develop a real-time teleoperated robotic system it is necessary to cope with the latency that is contributed by all the components of the system. Usually, the transmission elapsed time is one of the most serious issues in the latency problem. In the proposed teleoperated robotic system, there are a motion-capture system to acquire human motion data, a paddle-juggling robot to bounce a ping-pong ball up and down, and a proposed latency-removal algorithm to remove the motion latency between the human and the robot. It is assumed that the delay induced by the latency-removal algorithm is negligible. To reduce the lump-sum latency effect of the teleoperated system, the proposed latency-removal algorithm is based on a emulator/controller neural-network architecture to predict the robot motion command in advance. In the task of bouncing a ping-pong ball up and down, the movement is near cyclic. The simulation result shows that the performance by the proposed method is even better for a near cyclic movement.