Department of Validation Intelligence for Autonomous Software Systems

Autonomous Software Systems (AS) such as autonomous cars, cargo, or collaborative robots are characterized by their self-decision capabilities such as planning, scheduling and executing complex tasks with limited human intervention. Unlike automated systems, AS must be able to safely react to unexpected events like faults, hazards or unforeseen human behavior while still being able to complement human’s capacity to make decisions using vast amounts of uncertain data. However, their increasing relevance to address crucial societal challenges such as living in a more sustainable world with smart transportations and environment- and human-values preservation, has brought challenges for their validation and verification with respect to robustness, reliability, safety, and security.

To tackle these challenges and contribute to the wider adoption of trustworthy AI, the Validation Intelligence for Autonomous Software Systems Department (VIAS) explores three research leads which are:

  1. Using Artificial Intelligence to improve testing AS through foundational works on how to generate test suites that can be optimized, prioritized and scheduled on multiple test agents dedicated to AS using paradigms such as constraint optimization, constraint-based scheduling, and machine learning.
  2. Testing intelligent systems with adaptive capabilities raises new challenges, like handling the so-called oracle problem and the wide space of test scenarios for AS. The VIAS department explores the adaptation of established software testing methods, such as variability testing and metamorphic testing, as well as the creation of novel techniques towards the new challenges raised by intelligent systems.
  3. Artificial Intelligence for security by developing Intelligent methods to detect and correct software faults in AS. A focus of the VIAS department is the detection of falsified data injection in global surveillance systems in the maritime and transportation domain and the usage of machine learning to detect threats in AS.


Actively involved in European and National projects, the VIAS department nurtures a constellation of scientific and industrial collaborations with partners developing and maintaining AS in the domain of smart transportation and industrial robotics, where empirical evaluation and field experiments are crucial to demonstrate the potential of new fundamental research results.

Affiliation

Software Engineering

Contact person(s)