Using a distance metric to guide PSO algorithms for many-objective optimization

In this paper we propose to use a distance metric based on user-preferences to efficiently find solutions for many-objective problems. We use a particle swarm optimization (PSO) algorithm as a baseline to demonstrate the usefulness of this distance metric, though the metric can be used in conjunction with any evolutionary multi-objective (EMO) algorithm. Existing user-preference based EMO algorithms rely on the use of dominance comparisons to explore the search-space. Unfortunately, this is ineffective and computationally expensive for many-objective problems. In the proposed distance metric based PSO, particles update their positions and velocities according to their closeness to preferred regions in the objective-space, as specified by the decision maker. The proposed distance metric allows an EMO algorithm's search to be more effective especially for many-objective problems, and to be more focused on the preferred regions, saving substantial computational cost. We demonstrate how to use a distance metric with two user-preference based PSO algorithms, which implement the reference point and light beam search methods. These algorithms are compared to a user-preference based PSO algorithm relying on the conventional dominance comparisons. Experimental results suggest that the distance metric based algorithms are more effective and efficient especially for difficult many-objective problems