Sustainable Agent Communities
We believe that in the future information agents will represent their users interest in electronic marketplaces and other forums to trade, exchange, share, identify, and locate goods and services. Such information worlds will present unforeseen opportunities as well as challenges that can be best addressed by not individual agents but by agent communities. An agent community is a stable, adaptive group of self-interested agents that share common resources and must coordinate their efforts in utilizing and sustaining the group resources and structure. More specifically, agents will need mechanisms to benefit from complementary expertise in the group, pool together resources to meet new demands and exploit transient opportunities, negotiate fair settlements, exchange help and transfer knowledge between peers, monitor and secure the community against malicious intruders, and learn to work together effectively. In this talk, I will summarize some of our research results on trust-based computing, negotiation, collaboration and learning that will enable intelligent agents to develop and sustain vibrant, successful agent communities.
Han La Poutré
Action-Reaction in Multi-agent Games
Interactions between autonomous agents can be modeled as e.g. competitive games. Instances of competitive games are e.g. negotiations and auctions. Such games typically occur in a repeated fashion. In order to perform well in the complex settings of multiagent systems, agents have to learn during these repeated games. Computational intelligence techniques are then important techniques for learning agents.
The design of adaptive agents participating in repeated competitive games forms an important research subject. This research aims to be a fundamental answer to the important change of paradigm of ICT in society that is currently taking place: Many independent non-cooperating parties, that are connected by distributed networks, interact continuously in a dynamic and ad-hoc way; the terms for these interactions are set via negotiations and auctions, and the form of ad-hoc cooperation can be determined for each individual occasion.
Some issues that we address are the following:
- Design of adaptive agent strategies for competitive games in multiagent systems, like for negotiations, auctions, and dynamic pricing;
- Design of adaptive strategies where an agent takes into account the opponent's reaction to its own actions;
- Design of adaptive agents participating in repeated allocation mechanisms for e-business and logistics. Especially, taking into account future potential acquisition possibilities leads to advantages in an agent's market position. For the involved agents, however, it can also lead to a setting related to the multi-party prisoner's dilemma.
We also illustrate this with some recent research results in the application fields of distributed logistics (distributed transportion logistics, distributed patient logistics) and of e-commerce (sales of information services and pricing mechanisms).