TRACE - an adaptive organizational policy for multi agent systems

Proposes an adaptive organizational policy called TRACE (Task and Resource Allocation in a Computational Economy), incorporating task and resource allocation for multi-agent systems (MAS) that operate under time constraints and load variations. A MAS is comprised of several problem-solving organizations. Our task allocation protocol (TAP) takes requests and plans, and allocates subtasks to agents within an organization. As requests arrive arbitrarily, 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 our price-directed resource allocation protocol (RAP). Simulation results show that TRACE exhibits high performance despite unanticipated changes in the environment