A Framework for Dynamic Agent Organizations

This chapter presents an adaptive organizational policy for multi-agent systems called TRACE. TRACE allows a collection of multi-agent organizations to dynamically allocate tasks and resources between themselves in order to efficiently process and incoming stream of tasks. The tasks have deadlines and their arrival pattern changes over time. Hence, at any instant, some organizations could have surplus resources while others could become overloaded. In order to minimize the number of lost requests caused by an overload, the allocation of resources to organizations is changed dynamically by using ideas from microeconomics. We formally show that TRACE has the ability to adapt to load variations, reduce the number of lost requests, and allocate resources to computations on the basis of their criticality. Furthermore, although the solution generated by TRACE is not always Pareto-optimal, TRACE has the properties of feasibility and monotonicity that make it well suited to time-constrained applications. Finally, we present experimental results to demonstrate the performance of TRACE.