Abstract
It is important to prevent commuter trains overrunning which are caused by human errors. Although skillful drivers can robustly withstand drastic changes to situations, they may make a mistake. In this paper, we focus on a braking control problem in which the task of the human operator is to make the train stop exactly at a given position. We propose a support system for the human operator, which applies so-called Just-In-Time modeling to predictive control. Moreover, we propose a human-machine interface system which displays the level how far the current manipulation apart from the usual ones by using an input-output database. The effectiveness of the proposed method is demonstrated by simulations of a braking system on trains.
