Post-doctoral position on "formal approach to requirement engineering"

Context

Requirements engineering (RE) is critical in software and system design. Indeed, a major part of the cost of software and system development is known to be traceable to the understanding of the problem domain and requirements. In the last decade, much research on this topic has reached maturity and has been integrated into modeling methods.

Over the last few years, Onera has been developing a "core" language for requirements modelling, called  Khi, which allows to describe the behavioural goals of the system on the one hand, and the different agents of the system, with their behaviour, on the other hand. The main problem addressed by Khi is to assess possible assignations of agents to goals, i.e., to answer the following question: is a given set of  agents able to ensure a given goal?

In order to allow formal verification, Khi comes with a formal semantics relying on a temporal multi-agent  logic called Updatabe Strategy Logic (USL). USL generalizes temporal logics with the ability to reason about the strategies applied by agents in order to fulfill their objectives.

Although very promising, Khi lacks a tractable way to model and reason about the problem domain, i..e., the conceptual entities that form the system and its environment, as well as their types, relationships, etc. Although some approaches allow to model a structural view of the domain, a unified formal framework considering the problem domain, goals and agents altogether is still lacking.

Description of the activity

The aim of this postdoctoral position is to devise:

•  an extension of Khi allowing to reason about both the expected behaviour of the system and the main concepts that form the domain. The underlying formal semantics will build on USL on the one  hand, and on a formal framework well-suited for domain modelling (e.g., first-order logic, relation algebra, description logics or algebraic specification) on the other hand. The choice for the latter will result from a trade-off between expressivity, user-friendliness and the potential for automated  verification. Following an ongoing work concerning USL,  meta-theoretical properties of the proposed logic will be possibly  validated using a proof assistant such as Coq or Isabelle.
  

• a method, possibly supported by a software prototype, allowing to design and validate models of the above-mentioned Khi extension.
  

Application

This is a joint proposition between two Toulouse computer science labs:  Onera/DTIM and IRIT. Candidates are expected to be fluent in English or in French and should send a resume via email to Julien Brunel <[email protected]> and Jean-Paul Bodeveix <[email protected]>.

A complete version of this proposition is available here.