基于粒子群优化的有限反馈干扰对齐码本设计

Codebook Construction for Interference Alignment with Limited Feedback Based on Particle Swarm Optimization

大尺寸码本设计是有限反馈干扰对齐面临的关键问题之一,最优矢量量化码本设计问题可以等价为格拉斯曼线性装箱问题,只在某些特定条件下才存在解析解,通常采用计算机搜索或者信源编码中矢量量化算法求解,计算复杂度高,不利于大尺寸码本设计。从降低大尺寸码本设计计算复杂度出发,该文提出一种基于加速全面学习粒子群优化的码本设计算法。在全面学习粒子群优化算法实现简单,对非线性问题,特别是多峰值问题具有较强全局搜索能力的基础上,通过向最优粒子学习提高算法初期收敛速度,通过最大运动速度缩减提高算法后期收敛速度和算法性能。试验结果表明,无论是相对于经典粒子群优化算法和全面学习粒子群优化算法,还是相对于广义Lloyd算法,新算法均能在降低算法复杂度的同时提高算法性能。

Finding the optimal codebook is one of the key problems for interference alignment with limited feedback,it is equivalent to line packing issue in the Grassmannian manifold.Because analytical construction of the optimal codebook is possible only in very special cases,numerical search algorithms or generalized vector quantization algorithms for source coding are often sought to obtain near-optimal codebooks,but these algorithms characterize with poor performance and high complexity.In order to reduce the complexity of codebook construction,a new accelerative Comprehensive Learning Particle Swarm Optimization（CLPSO） algorithm is proposed.The convergence rate during the early period of the algorithm is speeded by studying of the best particle,the convergence rate during the later period is speeded and the performance of the algorithm is improved through reduction the maximum velocity of particles based on the CLPSO algorithm’s advantage of easy implementation,performing well on searching the optimal solution within defined space for non-linear problems,especial for complex multimodal problems.The simulation results show that the new algorithm achieves better performance than Particle Swarm Optimization（PSO）,CLPSO and Generalized Lloyd Algorithm（GLA） with low computational complexity.