Research On Memristor Array-Based Swarm Intelligence Algorithm

Memristor,as a new circuit element,has the characteristics of memory and nonlinearity,and can realize the integration of memory and computing.Memristor crossbar array,which is composed of a number of memristor,can realize more complex computing function and deal with harder problem compared to the single memristor.The emergence of memristor is expected to solve the bottleneck problem under the von Neumann system.Similar to the idea of memristor array,the swarm intelligence algorithm is also an example of intelligent group composed of multiple simple individuals,and some behaviors of memristors are similar to typical swarm intelligence algorithms represented by genetic algorithms and ant colony optimization algorithms,so memristor arrays are also used in the deployment of swarm intelligence algorithms.In this context,this thesis explores the similarity between memristor array and genetic algorithm and ant colony optimization algorithm.The main work and contributions are summarized below:1.Memristor array-based genetic algorithmIn this thesis,a matrix genetic algorithm is designed for the image edge detection task.The mapping relationship between memristor array and genetic algorithm is expounded,the image edge detection related operators of genetic algorithm are converted into matrix operations,and circuits based on memristor array are designed,which verify the correctness of the memristor array circuits through simulation.In the experiment,compared with the baseline memristor-based genetic algorithm,Sobel operator and Prewitt operator,the proposed algorithm improves the Precision score by 0.23,0.017 and 0.019,respectively,and the Recall score by 0.23,0.022 and 0.023,respectively,which proves the superiority of the proposed algorithm in image edge detection.2.Memristor array-based ant colony optimization algorithmsThe similarity between the behavior of ant colony optimization algorithm and the characteristics of memristor is discussed,the similarity between memconductance value and ant pheromone is elaborated from the two-way problem and multiple ways problem,the mapping relationship between memristor and ant colony optimization algorithm is summarized,and an ant colony optimization algorithm for two-dimensional path planning is designed.Experimental results prove the feasibility of deploying ant colony algorithm with memristors.Moreover,the shortest path length searched by the proposed algorithm is reduced by 25.6%,and the running time is shortened by 14.2%compared with the basic ant colony algorithm.3.Visualization system of typical swarm intelligence algorithmsBased on the proposed algorithms,combined with the ideas of software engineering,a visualization system of typical swarm intelligence algorithms is designed and implemented.The system provides user registration,login,image edge detection and 2D path planning functions for easy user use.