Accurate stopping technology that is used to stop a train precisely at a desired position has been continuously studied. Recently, this technology has become more important as platform screen doors have been installed in metro stations for passenger safety. Various algorithms have been proposed for accurate stopping. However, it requires a lot of time and money to verify experimentally the performance of novel algorithms with actual trains. Hence, it is difficult to improve and verify control performance with newly proposed algorithms. To resolve this problem, we propose to build a test bed for the accurate stopping control using scale models of train and track. After analysing the acceleration and deceleration characteristics of a specified train that is going to be equipped with a novel proposed algorithm, an adjuster is added to make the acceleration and deceleration characteristics of the scale model be similar to those of the target train. System identification was also performed by analysing the step response with various step input to the scale model. In this study, feedforward with proportional and integral (PI) controller was proposed, where gain parameters were determined based on the system characteristics analysed by experiments. The structure of the test bed and the process of building the test bed for the accurate stopping control are described in this paper. Experiment results are also provided to show the performance of the proposed controller. By being tested with this test bed, an improved control algorithm for the accurate stopping is expected to be derived.