The authors describe a new approach to multiagent resource allocation problems. A typical problem would be the allocation of resources in an emergency room situation as new cases arrive continually. The model consists of four components: tasks, task agents, resources, and resource proxy agents. The task agents act on behalf of the tasks and interact with the resource agents. The resource proxy agents listen to requests from the task agents and choose which requests to pass on to resources. All the resource proxy agents are known to the tasks. Task agents request resources, based upon a strategy generated by the task agent. The resource agents decide which tasks receive resources. This decision allows for preemption of existing allocations.

Task agents use two parameters in their determination of which resources to request. These are congestion (the level of competition for resources) and churn (the rate of change of the environment). The algorithm, which is described in some detail, contains an interesting feature, which is a notion that the authors call bother and is used to emulate scenarios such as humans who are too busy or annoyed to check the recommendations of the system. The authors describe the algorithm first at a high level and then in more detail. The high-level description allows the authors to present a cleaner description of the various components. Finally, the authors present comparisons with other systems that have been used for resource allocation. The gradualist narrative makes for a lucid description of the system. The paper is recommended to anyone interested in multiagent systems.