The AI for Cognitive Cyber-Physical Systems (AI4C2PS) initiative explores how artificial intelligence can enhance interoperability between cyber-physical systems (CPS) and humans.
Our research focuses on:
3 major challenges
Challenge 1: Digital Twin-Based Cognitive CPS
Problem:
How can a system of Cyber-Physical Systems (CPS) be sufficiently trained by Digital Twins simulations/emulations, so that it becomes fit for purpose to operate within the true physical realities of systems interoperability ?
Explanation:
CPS combine sensors, artificial intelligence, and real-world interactions. Digital Twins allow these systems to be tested and trained in a simulated environment before deployment. But how can we ensure that learning in a virtual world translates effectively to real-world performance? This is a key challenge.
Challenge 2: Explainable AI for humans and agents
Problem:
Can knowledge graphs and reasoning based on ontology concepts inferences provide explainability for AI processes in a way interpretable by both machines and humans, to support CPS-CPS and CPS-Human interoperability ?
Explanation:
AI often makes decisions that seem opaque (a « black box »). By integrating knowledge graphs, which organize information into logical relationships, and ontologies, which define concepts and their meanings, we can improve explainability. The goal is to enable both humans and machines to understand and interact effectively with AI.
Challenge 3: Human Digital Twin
Problem:
Can we build a Human Digital Twin based on semantic and stochastic models and supported by DRL and xAI, able to model human workers realistically enough to ensure proper interoperations between CPS and humans ?
Explanation:
The idea is to create a realistic digital model of a human that reflects their behavior and interactions with CPS. This could improve training, safety, and performance in automated environments.