基于模拟退火粒子群算法的认知引擎研究

认知引擎的基本功能之一就是根据复杂多变的无线环境及业务需求,利用多目标优化策略,自适应地调整无线参数,实现动态环境下的可靠通信。目前,很多研究的重点集中在遗传算法(GA)及其改进算法上,但其收敛速度较慢,不利于复杂多变以及实时性要求较高的系统。对此,提出一种模拟退火粒子群算法(SABPSO),它采用模拟退火与粒子群算法交替迭代的方式,协同寻优。其可有效提高收敛速度,并克服基本粒子群算法易于陷入局部极值的缺点,增强全局寻优能力。最后,在不同通信模式下,利用多载波系统进行仿真,结果表明,SABPSO算法在收敛速度和平均适应度上优于基本算法。

Cognitive engine research based on simulated annealing and particle swarm optimization algorithm

One of the basic functions of cognitive engine is to adjust adaptive wireless parameters and use multi-objective optimization strategies to achieve reliable communication in a dynamic environment according to complex wireless environments and service needs. Currently many studies focus on the genetic algorithm (GA) and its improvement, however, its slow convergence is not conducive to complex and high real-time systems. We propose an algorithm, called simulated annealing particle swarm optimization (SABPSO), which combines the simulated annealing algorithm and the particle swarm algorithm to search jointly for a better solution in an alternately iterative manner. It can effectively improve the convergence speed and overcome PSO's shortage of falling into local extreme values easily, thus enhancing the ability of global optimization. Finally, multi-carrier system simulation results in different communication scenarios show that the SABPSO outperforms the basic algorithms in convergence rate and average fitness.