Online Verification of Control Parameter Calculations in Communication Based Train Control System

Communication Based Train Control (CBTC) system is the state-ofthe-art train control system. In a CBTC system, to guarantee the safety of train operation, trains communicate with each other intensively and adjust their control modes autonomously by computing critical control parameters, e.g. velocity range, according to the information they get. As the correctness of the control parameters generated are critical to the safety of the system, a method to verify these parameters is a strong desire in the area of train control system. In this paper, we use our experience learned during verifying a CBTC system to present our ideas of how to model and verify the control parameter calculations in a CBTC system efficiently. – As the behavior of the system is highly nondeterministic, it is difficult to build and verify the complete behavior space model of the system offline in advance. Thus, we propose to model the system according to the ongoing behavior model induced by the control parameters. – As the parameters are generated online and updated very quickly, say every 500 milliseconds in the case we met, the verification result will be meaningless if it is given beyond the time bound, since by that time the model will be changed already. Thus, we propose a method to verify the existence of certain dangerous scenarios in the model online quickly. To demonstrate the feasibility of these proposed approaches, we present the composed linear hybrid automata with readable shared variables as a modeling language to model the control parameters calculation and give a path-oriented reachability analysis technique for the scenario-based verification of this model. We demonstrate the model built for the CBTC system, and show the performance of our technique in fast online verification. Last but not least, as CBTC system is a typical CPS system, we also give a short discussion of the potential directions for CPS verification in this paper.