7153 Copyright © 1996 IFAC 9b-Ol 3 13th Triennial World Congress, San Francisco, USA EARLY VALIDATION OF REAL-TIME SYSTEMS BY MODEL EXECUTION * Miguel A. de Miguel, Juan C. Dueiias, Alvaro Rendan *; Juan A. de la Puente, Alejandro Alonso, Gonzalo Lean E. T.S.I. Telecomunicaci6n, Universidad PoliUcnica de Madrid Ciudad Universitaria, sin, Madrid, E-28040 Spain Phone: +34 1 549 5700 x 381 E-mail: mmiguellOdit.upm.es • This work has been partially funded by the EEC ESPRIT programme under contract no. EP8593, and by the CICYT Spanish national programme as project TIC94-1542-CE . •• Alvaro Rend6n is visiting professor from Universidad del Cauca, Colombia, and his work is sponsored by Colombian Institute for Science and Technology Development (COL CIENCIA S ) and the IDERS project (ESPRIT EP8593). Abstract. The late validation of temporal aspects is still today the great bottleneck in embedded real-time systems (ERTS) development. In this article, a twofold approach is presented to reduce this problem: to enhance the visibility of temporal aspects in the product through the use of the proper design notations and models, and to integrate these notations with tools that allow the validation in a stage previous to coding. Both tools and notations allow the early feedback of functional and timing behavior, provide automatic support for the validation, and reduce the conceptual distance between the system specification and its implementation. Key Words. Real-time systems, executable design models, animation of prototypes, testing and validation 1. INTRODUCTION From a very general viewpoint, the main problems in the development of embedded real-time systems (ERTS) are related to: • the large conceptual gap between both specifica- tion and design techniques currently used, and the implementation and Real-Time Operating System (RTOS) usage, • temporal validation is only available on the full final system code. Differences between the development environment and the target architectures make the timing differ. Some approaches try to solve these problems from differ- ent viewpoints or for some specific domains. It is worth while to see their advantages and problems in order to understand our aims: ROOM (Selic et al., 1994) is a methodology based on object-oriented concepts, mainly applicable to specification and design of soft Real-Time Sys- tems (RTS). It supports the animation/execution of models for early validation, although its use of time is somehow restrictive, being handled only at the specification level (computation times are not considered). Validation is only performed by means of model animation. HRT-HOOD (Burns and Wellings, 1995) -also based on object-oriented concepts- has been introduced for the design of hard RTS. These being dependable systems, checking for the schedulability by analytic techniques is mandatory, and HRT-HOOD makes extensive use of Rate Monotonic Analysis (Klein et al., 1993). Time (deadlines and computation val- ues) is taken into account, but it can not offer for- mal support for animation or simulation of the user model. Statemate (Harel, 1990) supports the Statecharts for- malism. This is a wise approach to the development of reactive systems, focused on the specification of these models, specially well-fit for synchronous sys- tems and based on structured methods. This method lacks the capability for modeling re- sources in the system under development (SUD),