Speeding up CapBasED Activities in Delegation Driven Multi-Agent Information Systems

Eleanna Kafeza and Kamalakar Karlapalem
Department of Computer Science
Hong Kong University of Science and Technology
Clear Water Bay, Kowloon, Hong Kong
{kafeza,kamal}@cs.ust.hk

Many of the applications in Multi-Agent Information Systems (MAISs) are activities (set of interdependent tasks) composed by set of
sub-activities. A task is an atomic activity executed by a single agent, where as a sub-activity consists of one or more interdependent tasks that can be delegated to other agents. We undertake our work in the context of Capability-Based and Event Driven Activity Management System (CapBasED-AMS). In a delegation driven multi-agent information system, an agent delegates certain sub-activities to other agents. In such environments, it is very much required to provide scheduling facilities where activities can be speeded-up.  Although there is a significant amount of work on static and dynamic scheduling of activities (e.g. earliest deadline first, precedence constraints scheduling), in all of the existing work when a task is submitted for execution by the scheduler its order for execution is fixed. On the other hand, when real life agent-based information system applications are executing in a highly competitive environment; unpredictable events force a user/agent to request an acceleration of his/her activity. For such applications, the MAIS must have the ability to “adjust” and speed-up specific activities. In this paper, we model the execution of activities in a delegation driven MAIS. We formulate the speed-up problem and propose an algorithm that consistently and efficiently speeds-up activities. Thus, we look at the core system support required for speeding up activities in an MAIS.

Download Full Paper