The junior research group “Cyber-Physical Autonomous Systems” (CyPhy) aims at a new generation of reliable and trustworthy intelligent systems that seamlessly integrate with human environments. We unite the machine-centred and human-centred perspectives by modelling the entire human-AI interaction loop as dynamical systems. To drive these systems, our goal is to develop hybrid control strategies that combine established model-based (e.g., Model Predictive Control) and promising Machine Learning/data-driven (e.g., Deep Reinforcement Learning) approaches. We then apply these strategies to simulate, optimize, and implement intelligent systems with which humans can interact (more) naturally through new and more accessible user interfaces. With our application-driven research, we aim to bridge the gap between advanced theory and impactful, real-world solutions.

Projects

More about current and future projects of this group will follow.

Team

Lead

The junior research group leader, Dr. Arthur Fleig, joined ScaDS.AI Leipzig in March 2024. He studied Mathematics at the University of Bayreuth and completed his PhD on “Model Predictive Control of the Fokker-Planck equation” under Professor Lars Grüne’s supervision. Before joining ScaDS.AI, he worked as a postdoc in Professor Jörg Müller’s group in Bayreuth, where he applied his mathematical core expertise to the field of Human-Computer Interaction (HCI).

Since 2019, Fleig focuses on modelling, simulation, and optimal control of real-world-relevant dynamical systems. He develops and combines optimal control methods – such as Deep Reinforcement Learning and Model Predictive Control – with Human-Computer Interaction. In various international (UK, Norway, Finland) collaborative teams, he introduced the Optimal Feedback Control perspective to the HCI community. With OptiTrap, the first structural numerical approach to compute trap trajectories for acoustic levitation displays, he is one of a select group of HCI researchers to have published in the ACM Transactions on Graphics SIGGRAPH journal.

Fleig’s research passion lies in viewing interaction with intelligent systems from both a machine-centred and a human-centred perspective. This incorporates both the user-centred design approach as well as a simulation- and optimisation-based approach to modelling interaction with these systems. By considering all three elements of the human-AI interaction loop – the user, the interface, and the intelligent system – as dynamical systems, this group aims at a new generation of reliable and trustworthy intelligent systems with which humans can interact (more) naturally through new and more accessible user interfaces.

Team Members

Publications

So far, no publications exist.