A strategy for multi-agent resource allocation under incomplete information

This paper tackles a distributed resource allocation among selfish and myopic agents for the redistribution of indivisible resources without side payment. Agents have divergent preferences among bundles of resources. Particularly, we address a negotiation strategy allowing an allocation with a good quality measured in term of social welfare. The aim is to obtain an allocation with near optimal social welfare (instead of an optimal one) since we don't allow any money compensation. The negotiation model we use is one-to-one with resource swapping as a contract type which can involve more than one resource at a time. Moreover, the approach we propose involve without any privacy information revelation (preferences): an extreme case is addressed where an agent has to obtain a vital resource (present in more than one preferred bundles) without reveal its weak position. Experiments on large problem instances demonstrate the effectiveness of the approach.