Full name: Han La Poutré
Formal name: Prof.dr.ir. J.A. La Poutré
Function: Group leader, Scientific Staff Member, MT member

Email: Han.La.Poutre@cwi.nl
Telephone +31(0)20 592 4082
Room:  M375

Research group: (SEN4) Multi-agent and adaptive computation

Han La Poutré is head of the CWI research group Computational Intelligence and Multi-agent Games. He is also full professor at Eindhoven University of Technology, at the Capacity Group Information Systems. He was formerly affiliated with Princeton University, Utrecht University, and Leiden University.

Computational intelligence and multi-agent systems, as La Poutré explains, refers to adaptive systems for multiple parties that have to act together, plan together and solve certain problems together, but that can also compete against each other.

“In our group we try to develop computational technologies that possess a certain degree of intelligence,” he says. “Some of these technologies are nature-inspired, for instance evolutionary algorithms. They are based for instance on repeated selection and combination, which eventually leads to a new and better solution.” In practice, as he explains, this results in systems that can ‘learn’ and adapt to changes. “Imagine a collection of freight trucks that have to deliver certain products,” says La Poutré. “Which truck should pick up which product, where should it deliver it, and which route should it follow? An intelligent system can generate different solutions and calculate which one is most time and cost efficient. It will select the better solutions, combine them into new ones, and repeat this until an optimal solution emerges.”

La Poutré also focuses on multi-agent systems. An agent, in this case, is a software module that specifies a person’s or an institution’s interests and various tasks, for instance how certain goods are delivered, by whom, to whom and when. In a multi-agent system, several of such agents work together, for persons or institutions that all have their own interests. “We try to design methodologies to reconcile these interests,” says La Poutré. “We do this by developing agents that have to comply with certain ‘game rules’, like the rules in negotiations, auctions, or market mechanisms.”

Even more interesting, as La Poutré highlights, is the combination of computational intelligence and multi-agent systems. “In our example of the transport logistics,” he says, “this would result in a system that calculates optimal routes, but also takes into account things like supply and demand, forecasting, or trade mechanisms.” Other application opportunities are numerous, ranging from hospital planning to online advertisements, or distributed power generation in electricity networks. Some of the group’s findings, including for hospital planning, are already being implemented in practice.

Why this work is so interesting? “It is really challenging to design a system that does all of these things, and can be flexible as well. These systems are particularly needed in the face of decentralization. Different parties have to work together in a dynamic environment, and we help them do this in an optimal way. To me that is very satisfying.”

collapse