Research on Intelligent Optimization Algorithm and Its Application in Communication
|School||Hangzhou University of Electronic Science and Technology|
|Course||Communication and Information System|
|Keywords||Intelligent Optimization Algorithm Multi-user Detection Blind Equalization Quantum Genetic Algorithm Particle Swarm Optimization Shuffled Leap Frog Algorithm Immune Algorithm Clonal Algorithm Cooperative Evolutionary|
The intelligent optimization algorithms,which are inspired by the evolution mechanism of nature and biology, have been widely discussed and applied to many different problems which the traditional optimization can not solve. The dissertation focuses on the performance improvement of the intelligent optimization algorithms, which are the quantum genetic algorithm (QGA), particle swarm optimization (PSO), shuffled leap frog algorithm (SFLA), the immune algorithm (IA) and clonal algorithm (CA), and their application in multi-user detection and blind equalization.The optimal multi-user detection can effectively suppress multiple access interference (MAI), multi-path interference and reduce near-far effects, but its computational complexity is too high. So the suboptimal multi-user detection which has reasonable degree of computational complexity and performance close to optimal multi-user detection is the main research content in this paper. First, a modified quantum genetic algorithm (MQGA) is proposed, in which the new evolutionary strategy with niche is used to initialize quanta swarm and an adaptive strategy is used to update rotation angle. Then, the multi-user detector based on MQGA (MQGA-MUD) is presented; the simulation results show that MQGA-MUD has better performance than multi-user detector based on the genetic algorithm (GA-MUD) and quantum genetic algorithm (QGA-MUD). Second, a discrete shuffled frog leaping algorithm (DSFAL) is proposed. To increase the diversity of frog species and improve further performance of DSFAL, Hopfield neural network (HNN), immune algorithm (IA) and clonal algorithm are respectively embedded into DSFLA, the Hopfield neural network DSFAL (HDSFLA), immune DSFLA (IDSFLA) and clonal algorithm DSFLA (KDSFLA) are presented, which have faster convergence and better abilities of seeking the global optimal solution than DSFLA. Then four multi-user detection methods-DSFLA-MUD, HDSFA-MUD, IDSFLA–MUD and KDSFLA-MUD respectively using DSFLA, HDSFA, IDSFLA and KDSFLA are obtained. The simulation results show that DSFLA-MUD has similar performance with MQGA-MUD; HDSFA-MUD has better performance than DSFLA-MUD, IDSFLA–MUD and KDSFLA-MUD have similar performance and their performances are better than other proposed detectors in this paper.Blind equalization can effectively reduce inter-symbol interference and be widely used in digital communication system. Firstly, a sixth-second order normalized cumulant for blind equalization algorithm is firstly proved. Then, to improve the performance of continuous SFLA, the optimizing SFLA (SSFLA) and stretching SFLA (NSFLA) using a new update strategy are proposed and applied in blind equalization. And then two novel blind equalization algorithms (SSFLA-BEQ, NSFLA-BEQ) based on SSFLA and NSFLA are presneted. The simulation results show that the presented sixth-second order normalized cumulant criterion has better performance; the proposed SSFLA-BEQ, NSFLA-BEQ algorithms have better performance than blind equalization algorithms based on SFLA. Lastly, to improve the ability of PSO solving continuous optimization problems, a modified algorithm (MDPSO) based on double-swarm particle swarm optimization (DPSO) with center communication is brought. Combining with cooperative evolutionary theory, a cooperative evolutionary algorithm (PIDPSO) based on MDPSO using the probability of selection is proposed and applied in blind equalization. The simulation results show that the blind equalization algorithm based on PIDPSO has better performance than blind equalization algorithms based on DPSO and PSO.