Benefits of multi-model architecture-based development for railway applications Aurelien Bocquet, Christophe Gransart INRETS-LEOST, Villeneuve d’Ascq, France Abstract Software design is now a main part of railway research. To ensure sufficient usability and efficiency, software becomes more and more complex to design and implement; new railway applications must answer to requirements like system interoperability, heterogeneity management. To achieve these goals, software designers may use middlewares. In this paper, we will present different existing types of middlewares, and the benefits of using a multi-model architecture-based development to improve the efficiency and safety of railway applications. We will thus present our multi-model architecture, supplying a generic programming model and a combination of communication models. Introduction Current railway related research fields aim at providing more safety, efficiency and services to existing transportation systems. But more and more of projects in these fields have to use software components in order to supply complex and efficient functionalities. And if a single component may supplies some interesting services, several interoperable components supply more constructed and complex processing; that is the reason why railway research must design and use communicating software components. The main problem is to find the best way to make components communicate. Actually, many communication systems exist, from basic inter-process communication [1] to complete design- integrated communication middleware like CORBA [2], and designers have to choose the best system, considering the communication behavior they designed their components with, and the kind of context they will be deployed in. In this paper, we will first expose the specific problems related to designing and deploying railway applications. Second, we will describe the main existing middlewares currently used to make components communicate. Then we will present our multi-model approach, based on these previous problems and means. To validate this approach, we designed a multi-model system to prove the interest of our works on a real problem related to railway services. Our multi-model software infrastructure is a complete application and implementation of this approach. We will thus present and detail it. The current state of our work and first evaluations we made on it will follow, and we will present related works to position ourselves in the research field. Then, we will finally conclude on this subject. Railway applications related problems Railway context needs software applications, which are able to communicate without suffering of problems on the communication link. Indeed, the reliability and quality of this link cannot be established once for all. For railway applications, the communication link can be either GPRS, GSM-R, UMTS, WiMAX, or Satellite, but in every case, the quality and the availability of the signal may vary : gap between access points or antennas coverage, ability to switch lossless while handovers occur, reliefs or tunnels, and so … So on one hand, railway applications design needs a communication system able to dynamically adapt to context variations and changes, handling disconnections and change of signal quality. On the other hand, it needs enough efficiency to fill requirements of railway