Research areas and current projects
Our research is concerned with modeling complex (and often stochastic) search and optimization problems and solving them through computational heuristics. A special emphasis is given to distributed problems arising in social multi-agent systems, which cannot be optimized by pure economic price coordination. Imposing a solution by a central planner will often be rejected by the involved autonomous agents. Therefore, we need to design distributed mechanisms that provide incentives to participate to the individual agents without deviating too much from a pareto-efficient solution. We apply these methods to various domains, leading the following complementary areas of research.
Further advancements of nature-inspired processes (Simulated Annealing, Genetic Algorithms, Connectionist Models) for stochastic and dynamic problems as well as suitable parallelization of the methods for the application in distributed systems (e.g. Peer-to-Peer-/Multi-Agent Systems).
Vehicle Routing Problems
Consideration of more advanced vehicle routing models involving electric vehicles, autonomous drones (or droids), and stochastic travel times depending on time of day and weather in order to meet the customer specific time windows.
Yield Management and Planning of Internal Processes
Dynamic Pricing of (classical and electronic) services and optimization of the resulting processes that might span multiple locations. (in particular for services with resource complementarities)
Modeling and optimization of processes across locations. Coordination is understood as game theory mechanism design for several autonomous players (men and/or software agents).
Optimization of reciprocal selection decisions in partner matching (e.g., job placement-, dating- or social-network platforms).
Medical Data Mining and Forecasting
Automated evaluation of socio-demographic and clinical data to create forecasts of future patient numbers that are differentiated based on their medical indications. Optimal design of supply structures based on regional patient potential.