We design a NCS (Networked Control System) where the communication between sensors and controllers takes place over a USN (Ubiquitous Sensor Network). In order to measure time delays between sensors and controllers in real time, we design an algorithm to measure RTT (Round Trip Time) between USN nodes, and implement it into TinyOS of USN. By using the measured time delays, we construct the Smith predictor to compensate the time delays between sensors and controllers in real-time. For the real time experiment, we simulate the dynamic plant model, controller, and USN interface using Real-Time Windows Target provided in MATLAB. The USN interface in the Simulink model consists of serial ports, which connect the plant output and controller with USN nodes. The experiment results show that the time delays between sensors and controllers are precisely measured in real time; the Smith predictor appropriately compensates the time delays; and the stability is achieved in the closed-loop of the NCS.