Here you can find information about some of my current projects and members of my research group.
This research track is dedicated to tackling the challenge of guaranteeing the correctness of properties employed in the synthesis of reactive controllers. Our recent work concentrates on advancing theories and methodologies related to assumption refinement, and formulation of heuristics designed to steer the search process towards 'ideal' specifications. Furthermore, our research explores the dynamic aspects of adaptation and degradation of specifications, particularly when controllers are deployed in novel and evolving environments.
Collaborators:
Sebastian Uchitel, Imperial College, UK
Renzo Degiovanni,
Universidad Nacional de Río Cuarto, Argentina
Jeff Kramer, Imperial College, UK
Axel van Lamsweerde, UCL, Belgium
Relevant papers:
Much effort within the software engineering research community has been dedicated to developing mathematically founded, practically useful methods and tools that support the development of systems in which software is the primary decision-maker and safety guarantor. Nevertheless, many systems remain human-intensive. With increased reliance on human expertise, the potential for errors, resulting from flawed (human) logic, oversight and misunderstanding, increases. Why and how can such systems fail? What can we do to build more robust software that can function correctly in the presence of human decision-making's complexity and dynamicity?
We explore new means for designing, modelling and verifying human-intensive systems to increase their dependability and effectiveness in decision making. We focus on methods for reasoning about human expertise, judgments and feedback.
Collaborators:
Hana Chockler, King's College London, UK
Joe Halpern, Cornell, USA
Jessica Woodhams, University of Birmingham, UK.
Matthew Tonkin, University of Leicester, UK.
Relevant papers:
While significant efforts have been invested in testing and verifying machine learning systems, less has been focused on the modelling and verification of requirements for such systems.
The challenges are exacerbated when the machine learning systems
are based on data that is dynamic and evolving. This necessitates innovative techniques for tracing, verifying and updating the requirements for such systems.
Our interdisciplinary research explores
new methods to support the (semi-)formal elicitation, traceability, verification and repair of requirements of AI-enabled Systems.
Collaborators:
Catherine Hamilton-Giachritsis, University of Bath, UK
Tim Grant, Aston University, UK
Jessica Woodhams, University of Birmingham, UK.
Juliane Kloess, University of Edinburgh, UK.