基于禁忌搜索的CDMA多用户检测算法研究

基于禁忌搜索 (tabusearch ,TS)的原理 ,提出了两种实现码分多址 (CDMA)系统的多用户检测 (MUD)的算法。一种是通过合理地选择初始解、当前解的邻域以及禁忌搜索表 ,构造禁忌搜索检测方法 ;另外一种是采用禁忌学习神经网络 (TLNN) ,通过神经网络状态方程的迭代完成最优检测目标函数的全局寻优。通过分析以及对同步和异步情况的仿真表明 ,这两种方法均具有多项式的计算复杂度 ,而TLNN比TS复杂度更低 ;它们优于传统、多级、Hopfield神经网络等方法 ,对远近问题不敏感 ,并且具有与最佳检测方法 (OD)接近的误码率性能。

Multiuser Detection Algorithms Based on Tabu Search

Based on the principle of tabu search (TS), two approaches for the multiuser detection problem in the CDMA communication system are proposed. One approach is to construct the TS algorithm directly by choosing the initial solution, the neighborhood of current solution and the tabu list properly. The other is to use tabu learning neural network, and realize the global optimization of the objective function by iterating the state equation. It can be found that the two approaches both have polynomial computational complexities, and the TLNN is better than TS. Simulation results for the synchronous and asynchronous cases are provided to show that they are a near far resistant, superior to CD, MSD and HNN, and have a near optimal BER performance.