Agent-based simulation is highly valuable tool for modeling and simulation of complex distributed systems
that previously only could be tackled in a too abstract way for useful statements. Based on this paradigm variable structures, adaptive actors, multi-level feedback loops, spatial and non-spatial interactions, etc. can be formulated so that these properties can be analysed explicitly.
Additionally, multi-agent simulation is a quite inuitive way of modeling complex distributed systems consisting of autonomous actors - the gap between modelling paradigm and real-world is smaller than in every other paradigm.
- Methods and tools for multi-agent simulation: SeSAm.
My basic research
goal is to improve methods and tools for agent-based simulation for
SeSAm is the first step for adressing these goals.
- guiding the design and implementation process so that a systematic way of developing valid and useful multi-agent simulations replaces the try-and-error procedure currently mostly applied
- providing rapid prototyping tools for being able to make experiences "playing around" with different models of complex systems without great effort in implementing these simulations
- providing comfortable and high-level languages that allow also users without deep simulation programming experience to design and implement their simulation models - make the methods of multiagent simulation accessible for people that were not able to bridge the technology gap before.
- Application of multi-agent simulation in interdisciplinary cooperations
- biology together with the people from Chair of Sociobiology,
U Würzburg, we developed several agent-based simulations concerning sociality and
task allocation in social insects (funded by DFG SFB 554- 2000-2006). Currently, there is a cooperation with Dr. Anna Dornhaus (University of Arizona, Tucson) devoted to the analysis of task allocation performance in social insects and software agents.
- traffic simulation between 1999 and 2004 in German-Brazilian Exchange Projects (DLR/CNPq and
CAPES/Baviera), as a cooperation with Profa. Ana L. Bazzan. This cooperation is continued and will hopefully be funded again in 2006 together with Prof. Kai Nagel (TU Berlin). Additionally there
is a SVI-funded project running until mid of 2007 with Emch + Berger AG, Bern for developing
an agent-based model for combined route and transportation mode choice.
- pedestrian simulation is one of our largest SeSAm application. March-Dec 2006, we simulated the pedestrian behavior of the SBB railway station in Bern based during the busy morning hours on real-world scale - complete layout, complete train schedules, all travelers according to SBB data. Our agents are intelligent and thus capable for flexible and individual path planning... This project was a coopertion with Emch&Berger AG Bern and financed by SBB.
- logistics we are simulating highbay warehouses in sufficient detail for providing
a useful basis for testing complex control software before the actual hardware is available. This is our most successful industry project. The SeSAm-models developed in this project are meanwhile applied
not only for software tests, but also for requirements engineering, user interface tutorials, etc... Project partner is SSI Schäfer Noell AG, Giebelstadt.
- health care and logistics. Our group had projects in the DFG SPP 1083 (project partner Prof. Dr. Kirn, U Hohenheim). The focus was on the simulation-based evaluation of different strategies for examination and therapy scheduling using agent-based concepts.
- there are several "minor" application projects, I'm involved in - they reach from shopping simulation, innovation dissemination in water economy, species diversity...
I'm generelly interested in multi-agent systems and their potential due to intuitive ways of representing
complex distributed systems that can be found in real world, but also artificial ones.
- Biologically inspired coordination in multi-agent systems (DFG SPP
554 - 2003-2006)
- Agent-oriented software engineering
How to make modelling and simulation more accessible to people without deep programming experience, needs knowledge about HCI. In particular I'm focussing on
- Simple, Efficient and scalable Visual Programming for multi-agent systems and simulations
- Interactions between humans and multi-agent systems for validating/falsifying implementation correctness, but also for generating inspiration for model building.