A Multi-Agent System for Improving the Resource Allocation on Programmes in Higher Education

A Multi-Agent System for Improving the Resource Allocation on Programmes in Higher Education

Constanta-Nicoleta Bodea (Academy of Economic Studies, Romania) and Radu-Ioan Mogos (Academy of Economic Studies, Romania)
Copyright: © 2013 |Pages: 23
DOI: 10.4018/978-1-4666-4038-2.ch005

Abstract

The chapter presents UNIRA, a multi-agent system developed by the authors for a Romanian university in order to improve the resources allocation for educational programmes and courses. Different types of resources are required to deliver programmes and courses. Considering a set of resources inquires, issued by programmes and courses, UNIRA system performs a transparent negotiation process between the managers of these resources, to find the solution for the allocation problem. During the initial stage of the multi-agent system deployment, only two types of resources are considered, professors and teaching rooms. The system is now in the validation phase. After the complete validation, the system will be integrated into the university management system. The UNIRA experience is relevant not exclusively for the academic resources management, but also for a large variety of domains, including the load distribution, production planning, computer scheduling, portfolio selection, and apportionment.
Chapter Preview
Top

Resources Allocation Problem And The Solving Approaches

The resource allocation problem (RAP) is how to allocate available resources to the clients (scheduled tasks or agents, either cooperative or self-interested) in a way which maximizes the global utility (Dolgov & Durfee, 2006). RAP is relevant in many domains, such as: sociology and marketing (for ranking different objects according to consumers' economic social preferences), product design, architecture and construction, operations, network routing, transportation logistics, bandwidth allocation and commercial transactions, just to name a few. The real-world applications usually work in complex environments, with high dimensionality, dynamic, non-cooperative and uncertain, which make RAP even more complex and difficult.

RAP is expressed as a Constraint Satisfaction Problem (CSP). RAP can be easily mapped into a list-coloring problem in a special kind of graphs called interval graphs. In this case, RAP is called restricted coloring or feasible coloring problem. It is known that the usual graph-coloring problem in interval graphs is linear, but the list coloring is NP-complete (Choueiry & Faltings, 1994). As an optimization problem, RAP is a multi-objective and over-constrained problem. Even if the optimality is rarely needed in the real-world situations, the optimal solution is expected by the users. For solving RAP, a centralized or distributed approach could be undertaken. A Distributed Constraint Optimization Problem (DCOP) is more adequate in a large problem space. In this case, the problem is split into agents, each of them having a specific set of variables and constraints as well as local optimization criterion. The goal is to find a feasible solution with the highest ranking by all agents (Ridder, Brett, & Signori, 2012).

Complete Chapter List

Search this Book:
Reset