PhD Student - FU Berlin

Working on agent-based fish models, using machine learning to bridge the gap between the swarm behavior and the individual behavior. Observer bias is avoided, by training the behavior directly from recordings of real fish. An additional topic is extending attribution methods to explain the models on different time- and swarm scales.
Affiliations
Sponsorships
Interests
Besides the PhD, I work on the energy transition and on mobility, usually with an optimization problem at its core:
- Power grid - how electricity is generated, transported and distributed, and how the grid keeps supply and demand in balance as more and more of it comes from wind and sun.
- Optimization and machine learning - good decisions from noisy measurements, planning under uncertainty.
- Collective behavior - agent-based models of fish swarms (PhD).
- Traffic systems - especially the alternatives to the car: cycling, public transport, and the data it takes to plan for them.
Projects
- efficharge - Charge any battery when electricity is cheap and green. A smart plug switches the charger on in the best window before you need it. Free web app, open source.
- velosense - How rough is every street for cycling, separately for the road, the cycle path and the sidewalk? Estimated from OpenStreetMap and bike rides recorded by phone, with bike routes on top.
Repositories
- robofish.io - Store and use swarm data in a unified way. The documentation is available here
- goban-svg - Create vector graphics of go boards.
Publications
- Fish adapt and dynamically avoid an
approaching robotic fish across repeated exposures
S. Van Havermaet, A. Gerken, D. Mazrekaj, D. Bierbach, P. Simoens, T. Landgraf, Y. Khaluf. Scientific Reports, 2026. - Robots that learn to evaluate models of collective
behavior
M. Hocke, A. Gerken, D. Bierbach, J. Krause, T. Landgraf. Preprint, arXiv, 2026. - Continuous Value Iteration (CVI)
Reinforcement Learning and Imaginary Experience Replay (IER) For Learning Multi-Goal, Continuous
Action and State Space Controllers
A. Gerken, M. Spranger. IEEE International Conference on Robotics and Automation (ICRA), 2019. - Behavioral diversity through homeokinesis in a compliant robot
A. Gerken, O. Berthold, V. Hafner. Poster, IEEE ICDL-EpiRob, Lisbon, 2017.
Contact
| Email: | info@agerken.de |
| GitHub: | AndreasGerken |