Certification of AI models

Applications of Artificial Intelligence have made inroads into several areas of society. In particular, this applies to safety-critical domains such as autonomous vehicles, energy systems and the medical sector. This will lead to an increased demand for testing and verification of AI systems. To meet this need, RISE develops methods for machine-assisted verification of AI models with respect to both security and functionality of models.

Application

Applications range from computer vision in autonomous vehicles to control systems and language models. In this project, we examine requirements for control systems with an AI model component that controls for instance a machine, robot or manufacturing process. Verification can be used to ensure that the control module is robust and stable during normal as well as extraordinary scenarios.

Goals

RISE develops prototyping tools for verification of AI models to be used in combination with available open source software. The methodology is validated in the project through a number of case studies. With the project, we bring verification of AI models from academic research closer to industry, business and society and reinforce our competence in the area.