It's expected that near future's earthquake disasters are by huge pulse-type ground motions of an inland earthquake or long-period type ground motions of a subduction zone earthquake. Some concerns about seismic-isolated buildings can be also pointed out as follows: 1) Extra-large deformation of isolators exceeding the design criteria due to those types of earthquakes, 2) Degradation of seismic-isolated effects for reduction of acceleration response reduction due to the continuation of the long-time vibration. To overcome those problems, authors focus on introducing semi-active control systems on isolation layer. In this study, real time hybrid tests are performed on the mid-story isolated structures with semi-active device using the rotary inertia mass damper encapsulating magneto-rheological fluid (MR rotary inertia mass damper). In this test, isolation layer and superstructure are actually created as the test specimen, and lower part of structure is simulated by lumped mass system model in a computer. By using actual equipment of damper which has not been exactly clarified its behavior, modeling errors seem to be avoided, and the accuracy of the results on the hybrid test can be secured. This paper reports about constructing the system for real time hybrid test, and verifying its validity, moreover verifying the reproducibility of shaking table to confirm the reliability of this test. At first, mechanical specification of the experimental system and devices are evaluated. Steady vibration tests by sinusoidal waves are carried out on MR rotary inertia mass damper, and then its mechanical properties are modeled for numerical evaluations. The modeled formula is confirmed to agree to the experimental value, and is considered reasonable. A control time lag is observed when semi-active control is performed. This time lag is considered on the first-order delay system, and is also evaluated from experimental data. By considering this time lag effects on the numerical simulations, the precise experimental results can be gained. Lots of cases of the real time hybrid experiments for various types of earthquake input motions are operated. Those results are also compared with the numerical results by the fully modeled analyses, and the validity of the testing system, the modeling of this damper and the identification of the specimen are confirmed. Moreover, performance and the reproducibility of the real time simulators, which is actualized as the displacement of the shaking table, are evaluated. As results, high reproducibility of displacement of the shaking table by real time calculation could be observed and controlling errors could be manipulated to be negligible. At the same time, generated accelerations on the shaking table are observed performing accurately. However there is a little error in the reproduced acceleration for some cases. As a concluding remark, it is confirmed that reliable experimental system to operate real time hybrid test is constructed through this research.