Coevolutionary Multi-Objective Discrete Particle Swarm Optimization for Gateway Placement Optimization Problem

초록

The gateways placement optimization (GPO) problem is crucial for the construction of low earth orbit (LEO) satellite networks. Previous research on GPO usually modeled it as a mixed-integer programming problem with single objective and simulated experimental data. This paper introduces a multi- objective combinatorial optimization model with multiple objectives and constraints for GPO. Furthermore, to efficiently solve for the non-dominated solutions, this paper proposes a novel algorithm named coevolutionary multi-objective discrete particle swarm optimization (CMDPSO). First, by introducing “multiple populations for multiple objective” framework, two swarms are initialized to optimize the numbers of activated gateways and relative deviation of traffic load, respectively. Second, bi-velocity mechanism is adopted to enhance its global search ability. Third, real global population distribution and internet exchange points data are utilized within the gravity model to estimate the traffic matrix. Last, in terms of the efficiency of the new algorithm, CMDPSO significantly outperforms several common multi- objective optimization algorithms, such as NSGAⅡ on GPO model.

키워드