Typical Swarm Intelligence Algorithm And Its Simulation Implementation On FPGA

Swarm intelligence is a complex system composed of many individuals with certain structure,function and behavior,which is usually manifested by the ability of groups to cooperate with each other and jointly complete some tasks,typical swarm intelligence optimization algorithm such as Ant Colony Optimization(ACO)and Particle Swarm Op-timization(PSO).On the one hand,the research of swarm intelligence is to modify al-gorithms in specific application scenarios.On the other hand,due to the inherent paral-lelism of swarm intelligence optimization algorithms,and the running speed of hardware is higher than that of software,it is worth exploring to realize swarm intelligence algo-rithms by using FPGA’s high flexibility programmability,short development cycle and high parallel computing efficiency.Based on the above elaboration,this thesis focuses on design and implementation of ACO and PSO in FPGA.The main contributions are shown as follows.(1)Bayesian-based Ant Colony Optimization AlgorithmIn the application of image edge detection,the ACO algorithm combined with Bayes theorem can effectively improve the efficiency of pheromone.In order to avoid the ACO algorithm falling into the stagnation state prematurely,random disturbance and adaptive pheromone influence factors are introduced in this thesis.After the global pheromone up-date was completed,random disturbance is added to the pheromone matrix after the update in order to increase the randomness of ant movement,and linear increasing pheromone in-fluence factor is used to replace the original fixed value.Experimental results show that the proposed algorithm performs better in noise and continuity.Compared with four clas-sical edge detection operators and ACO algorithm without Bayes theorem,the precision of the proposed algorithm is improved by 35.05 and 0.43 percentage points on average,and the recall is improved by 34.93 and 0.54.(2)OTSU-based Particle Swarm Optimization AlgorithmIn the application of image segmentation,the PSO algorithm is combined with two-dimensional Otsu method,which can effectively improve the running speed of the algo-rithm.In order to overcome the shortcoming of PSO algorithm which is easy to fall into local extremum,this thesis introduces the nonlinear weight of PSO and a new position update strategy.Nonlinear weight is used to replace the original linear weight.When the fitness value does not change for a long time,Levy flight is added to increase the ran-domness of the position.The results of 1000 experiments show that after adding the two strategies,the number of times falling into local optimal is reduced from 69 to 20 times,which can effectively reduce the probability of the algorithm falling into local optimal.(3)FPGA implementation of the two typical swarm intelligence optimization algo-rithmAiming at two typical swarm intelligence optimization algorithms,this thesis mod-ifies them according to the actual situation,and gives the design scheme of their FPGA implementation.Through the hierarchical analysis of the algorithm,each module of the algorithm is divided in a top-down manner.For each module of the algorithm,detailed functional design and specific implementation are carried out,and test files were written to pass the simulation verification in Modelsim.Experimental results show that it is feasible to combine typical swarm intelligence optimization algorithm with FPGA.