Towards Automotive Embedded Systems with Self-X Properties

Since the first pieces of software have been introduced into automobiles in 1976, the complexity of automotive software systems is growing rapidly. Today automotive software is widely installed for diverse applications ranging from the infotainment domain (e.g. entertainment, navigation, etc.) with typically no real-time requirements to safety-critical control software (e.g. engine control, safety functionalities, etc.) with hard real-time requirements. In addition, many comfort functionalities of automobiles are realized by software nowadays (e.g. the control of the air condition system, electronic window regulator, etc.). Up to 90% of today’s innovations in the automotive industry are realized by hardand software (Pretschner et al., 2007). This results in up to 2,500 ”atomic” functions realized in software on up to 67 electronic control units (ECUs) in modern high-end cars (Furst, 2010). For the future development of automobile electronics, there are two major trends: A growing number of functionalities and through this a growing importance of software in the car (Hardung et al., 2004). Future generations of cars will be equipped with many new, complex features (Czarnecki & Eisenecker, 2000). For example, functionalities to support active driving safety (e.g. driver assistance systems), features which enable new innovative driving concepts (e.g. engine control for hybrid vehicles), or new functionalities in the comfort domain (e.g. new infotainment features). Most of these functionalities will be realized in software, which increases the amount and importance of software within the automotive domain necessarily. But these new features will also increase the complexity of future vehicular system architectures. For instance, driver assistance systems increase the complexity because they interact with several in-vehicle domains, e.g. the power-train and infotainment domain. In future, the trend of establishing more and more interactions between software components will continue, e.g. through x-by-wire features, where mechanical transmission is replaced by electrical signals. This results in a growing interdependency of separated software domains and in an increased need for interconnection. Another important aspect is the continuously growing number of functional variants caused by customer-specific equipment options or country-specific regulations. At the same time, the demand on the software quality within the automotive domain is very high at all times. These requirements must be satisfied in the future, despite the increasing complexity of automotive software architectures. Even today it is a great challenge to manage these systems from the outside.