Cognitive radio resource scheduling using an adaptive multiobjective evolutionary algorithm

With the proliferation of IoT devices and the increasing popularity of location-oriented services in cyber-physical-social systems, the cognitive engines of these systems have taken on a multitude of parameters across various dimensions, making it impractical and time-consuming to search for the exact optimal solution. To address this challenge, the use of nature-inspired or evolutionary algorithms to find satisfactory solutions in a timely manner has gained significant attention, with reference point-based algorithms being one of the prominent approaches. However, when dealing with nonuniform, degenerate, and discrete Pareto fronts in the target space, using a considerable number of reference points may become ineffective, leading to a loss of diversity in exploration and exploitation during the problem-solving process. Consequently, the distribution of the solutions is adversely affected. To overcome this challenge, this paper presents a strategy to estimate the eigenvalues of the Pareto front in a timely manner. When encountering nonuniform, degenerate, and discrete Pareto fronts, a combination of radial space partitioning and angle selection mechanisms is employed to address these issues. Subsequently, an adaptive selection-based many-objective evolutionary algorithm (ASMaOEA) is proposed. Extensive comparisons with several competing methods on 31 representative benchmark problems demonstrate that ASMaOEA can provide a flexible configuration for decision engines in three typical scenarios involving cyber-physical-social systems. Furthermore, the analysis confirms that ASMaOEA can reduce the bit error rate and improve the system's throughput, thereby offering substantial benefits to the overall performance of the system.