A Framework to Evaluate Multi-Objective Optimization Algorithms in Multi-Agent Negotiations

Multi-objective optimization algorithms are designed to find Pareto frontier set. This set plays a major role in multi-agent systems' negotiations. Different applications might be interested in different parts of Pareto frontier. In this paper we present a framework to show how a multi-objective optimization algorithm is evaluated against others. We used eleven algorithms implemented in MOMHLib++ library to test our framework on a two agent negotiation of binary issues and binary dependency. But our framework is easily expandable to higher number of objectives and all types of negotiations. Our analysis shows that a single scalarization value of Pareto frontier is not enough to compare multi-objective optimization algorithms, as it is done in most cases.