Cycles "les défis scientifiques du 21e siècle"
Joseph Sifakis to give a lecture on embedded systems
Joseph Sifakis - @ CNRS Bellevue
On Tuesday, 10 March, Joseph Sifakis, winner of the 2007 Turing Prize, research director at CNRS and holder of the Schneider-Inria-Digiteo industrial chair, will give a lecture at the Academy of Science on the topic of embedded systems – new challenges in computer science, which Inria has made one of its priorities.
As part of the lecture series, “Research Challenges of the 21st Century", organised by the Academy of Science, Joseph Sifakis will discuss the problems and challenges facing computer science in seeking to improve embedded systems.
Embedded systems are small components but are becoming ever more a part of many objects in our daily lives. They are found in cars, vacuum cleaners, rockets and satellites. They perform a simple, precise task: ensuring that their host apparatus works properly. As they multiply in number and uses, they are having to meet requirements related to space allowances, energy consumption and especially robustness. The slightest failure in any of these respects can have disastrous consequences ranging from the breakdown of cars to the explosion of rockets. To deal with these constraints, researchers are pursuing work to improve their knowledge of embedded systems. Their studies involve the design of tools for analysis that combine both physical and computational methods.
Joseph Sifakis has participated in the development of a methodology for the enumeration and verification of computational models. The nature and quality of his work earned him the Turing Prize in 2007. Based on his experience in this field, he will present as part of his lecture the strengths and weaknesses of current models in the interest of improving them. As a reminder, Joseph Sifakis is a research director at the CNRS, also directs the Verimag laboratory and has held the Schneider-Inria-Digiteo industrial chair since 2008.
"Embedded Systems – New Research Challenges for Computer Science"
by Joseph Sifakis
Embedded systems are components that combine software and hardware and ensure vital operating features. They are characterised by continuous interaction with their physical environment. Their applications include many fields such as transportation, telecommunications, energy distribution and use in electrical and electronic products.
We currently do not have a theoretical framework that allows us to take into account both functional and extra-functional requirements during the design of embedded systems.
In order to understand and study the behaviour of these systems, one must be able to combine engineering methods based on physics and the methods of computer science. As a matter of fact, Computer Science draws on discrete computational models that are independent of time and physical resources.
Extending Computer Science by incorporating into it the paradigms of disciplines based on classical physics today constitutes a major challenge for computer science research.
Four aspects of this challenge will be discussed:
- unifying analytical models and discrete computational models.
- building systems by combining components.
- ensuring system features are built-in in order to overcome the complexity and current limitations on post-facto verification.
- ensuring the predictability of interactive systems behaviour in order to overcome the uncertainty inherent in their dynamic behaviour.
In conclusion, a few fundamental differences between the two disciplines and their consequences will be discussed.
T.A. Henzinger and J. Sifakis. The Discipline of Embedded Systems Design Computer, October 2007, pp. 32-40.