The sub-project QAI2C (Quantum AI for Coordination and Classification) conducts fundamental research into the possibilities and potential of using quantum computing for selected mechanisms of coordination and classification in artificial intelligence (AI). The aim is to develop new quantum-assisted AI methods for agent-based coordination of actions in rational coalitions and through deep reinforcement multi-agent learning and deep unsupervised learning for classification, which can be more efficient than selected classical reference methods. The research work is being carried out with reference to the general problems addressed in the consortium, namely the application cases of capacitive vehicle route planning, chassis design and quality assessment of finite element meshes in component development by partners TWT and Mercedes-Benz. To investigate potential quantum benefits, a theoretical and experimental analysis of the developed QKI methods will be carried out using simulation on classical computers as well as on real (NISQ) quantum computing infrastructure available within the consortium, in comparison with the respective classical reference methods. Methods and data will be made available in an open software library for the overall QAIAC project. At the end of the QAI2C sub-project, we will have initial or more detailed findings on the adaptation of selected classical AI methods for coordination and classification to quantum computing. This also includes a broadened base of prototypically implemented quantum AI methods, as well as initial insights into the extent to which and under what conditions these could offer improved performance for the applications considered in the consortium.
Partners
- Mercedes-Benz
- FZ Jülich
- TWT
- E.on
