Section: Partnerships and Cooperations

Regional Initiatives

CASERM (Persyval-Lab project)

Participants : Pascal Fradet, Alain Girault, Gregor Goessler, Xiaojie Guo, Maxime Lesourd, Xavier Nicollin, Stephan Plassart, Sophie Quinton, Jean-Bernard Stefani, Martin Vassor.

The Caserm project represents a significant effort towards a Coq -based design method for reconfigurable multi-view embedded systems, in order to formalize the structure and behavior of systems and to prove their main properties. The use of a proof assistant to support such a framework is motivated by the fact that the targeted systems are both extremely complex and critical. The challenges addressed are threefold:

1. to model software architectures for embedded systems taking into account their dynamicity and multiple constraints (functional as well as non functional);

2. to propose novel scheduling techniques for dynamically reconfiguring embedded systems; and

3. to advance the state of the art in automated proving for such systems.

The objectives of Caserm that address these challenges are organized in three tasks. They consist respectively in designing an architecture description framework based on a process calculus, in proposing online optimization methods for dynamic reconfiguration systems (this is the topic of Stephan Plassart's PhD), and in developing a formal framework for real-time analysis in the Coq proof assistant (this is the topic of Xiaojie Guo's and Maxime Lesourd's PhD).

The Caserm consortium gathers researchers from the LIG and Verimag laboratories who are reknowned specialists in these fields. The project started in November 2016 and was completed in November 2019.

SEC: Construction of Safe Explainable Cyber-physical systems

Participants : Gregor Goessler, Thomas Mari.

In cyber-physical systems (CPS), software interacts with physical processes so as achieve desired functionalities. CPS are usually subject to safety and reliability requirements. Depending on the application, their failure may have unacceptable consequences, it is therefore crucial to ensure their correctness at design time. In addition, explainability of increasingly autonomous CPS is becoming crucial in order for the CPS to be socially acceptable.

The goal of this project is twofold. First, we will investigate a contract-based design approach for safe CPS in which different aspects – such as functional requirements, real-time constraints, and continuous behaviors – are modeled and verified separately. Second, we will leverage the contracts in order to ensure explainability of the system behavior by construction. By explainability we understand, informally, that for any behavior of the system we can automatically construct, from a log generated by the execution, an excerpt that retains only the events that causally contributed to the outcome, and that is easy to understand by a human expert.

The SEC project is supported by the “Initiatives de Recherche Stratégiques (IRS)” program of the IDEX UGA. It funds the PhD thesis of Thomas Mari, who will be co-advised by Gregor Gössler and Thao Dang (Verimag ).