G. Rudolph et al. (Eds.): PPSN X, LNCS 5199, pp. 773–783, 2008.
© Springer-Verlag Berlin Heidelberg 2008
Team Algorithms Based on Ant Colony Optimization –
A New Multi-Objective Optimization Approach
Christian Lezcano, Diego Pinto, and Benjamín Barán
Polytechnical School - National University of Asunción
P.O. Box 2111 - Paraguay
{clezcano,dpinto,bbaran}@pol.una.py
http://www.una.py
Abstract. This paper proposes a novel Team Algorithm (TA) approach based
on Ant Colony Optimization (ACO) for multi-objective optimization prob-
lems. The proposed method has shown a significant cooperative effect of dif-
ferent algorithms combined in a team of algorithms, achieving robustness in
the resolution of a set of various combinatorial problems. Experimentally, the
proposed approach has verified a balance on different performance measures
in problems as the Traveling Salesman Problem (TSP), the Quadratic Assign-
ment Problem (QAP) and the Vehicle Routing Problem with Time Windows
(VRPTW). Robustness and balance are achieved due to a novel classification
and selection of the algorithms to be used by the team, considering Pareto
concept.
Keywords: Team Algorithms (TA), Ant Colony Optimization (ACO) and
Multi-objective Optimization Problem (MOP).
1 Introduction
A Multi-objective Optimization Problem (MOP) can be defined as the problem of
finding a set of solutions that satisfies given constraints and optimizes several objec-
tive functions simultaneously. Usually, these objective functions are in conflict with
each other [1]. MOP is widely treated with different optimization paradigms [1-4], as
Multi-Objective Evolutionary Algorithm – MOEA, which have been successfully
applied to solve highly complex problems. On the other hand, the "No Free Lunch -
NFL" theorem [5] states that on average, all algorithms have the same performance.
Considering the NFL theorem, the development of Team Algorithms (TA) [6] is a
valid alternative for achieving high robustness on average.
This paper proposes a novel TA approach based on Multi-Objective Ant Colony
Optimization (MOACO) algorithms for solving combinatorial optimization problems
in a multi-objective context. Basically, the new proposal tries to improve the state of
the art in this area [6]. Experimental tests were carried out on different bi-objective
instances of the Traveling Salesman Problem (TSP), the Quadratic Assignment Prob-
lem (QAP) and the Vehicle Routing Problem with Time Windows (VRPTW). To