Requirements-driven Synthesis of Optimal Strategies for Autonomous Systems

Day - Time: 23 November 2017, h.14:30
Place: Area della Ricerca CNR di Pisa - Room: C-29
Speakers
  • Yehia el Rakaiby (University of Limerick, Lero)
Referent

Alessio Ferrari

Abstract

Many software systems have to operate in uncertain and dynamic environments, which often makes it difficult to model or specify their ideal behavior at design time. Thus, it is much more convenient to follow a requirements-driven approach: prescribe the software system's high-level requirements and provide it with the capabilities that enable it to decide autonomously of its optimal course of actions based on its requirements at runtime. In this talk, we will highlight some of the challenges of autonomous systems, discuss the limitations of some of current state-of-the-art approaches and then present REact, a requirements-driven approach for the synthesis of optimal strategies for autonomous systems. The main novelty of REact is that it enables the computation of optimal strategies based on requirements which can be identified using well-established goal modeling techniques. It also relies on a sound mathematical model, namely Markov Decision Process, that guarantees the optimality of the computed strategies.