The Research Of Multi-objective Circuits Evolutionary Design Based On Matrix Coding CGP

As a new development field,evolution hardware can automatically adjust the internal structure of the circuit according to the changes of the external environment,obtain the target circuit through evolution,and realize the self-organization,self-repair and self-adaptation of the circuit.The development of this technology will be able to solve problems such as circuit design and troubleshooting in harsh environments where space,strong radiation,deep-sea and strong viruses are difficult for humans to reach and difficult to repair.Evolutionary circuit design is one of the research directions of evolutionary hardware,and it is the focus of current evolutionary hardware technology research.The research of circuit evolution design mainly includes two parts:genetic algorithm and programmable logic device.Based on the Cartesian genetic programming,when using the one-dimensional vector coding scheme for multi-objective circuit evolution design,problems such as premature convergence and slow convergence rate are easy to occur.This paper adopts two-dimensional matrix coding scheme and a stepwise dimension reduction modular decomposition approach to conduct circuit evolution design research.This article has developed the following research around this method:(1)In the(?(10)?)evolutionary strategy,the relationship between the parameters such as the size of the population,the mutation rateand the convergence rate is studied.The results show that when the population size is less than 5,the evolutionary generation decreases rapidly with the increase of the population size.When the population size is greater than 5,the evolution of population size increases,and the evolutional generation decreases slowly.When the mutation rate_cP(28).006around,there are fewer evolutionary generations.(2)For one-dimensional vector coding schemes and two-dimensional matrix coding schemes,in this paper,we study the relationship between the levels-back and the speed of convergence and the activation of nodes;the relationship between the number of programmable logic cells and convergence speed and the activation of nodes.The results prove that,with the increase of levels-back,the more convergence speed;The number of programmable logic cells are appropriate,the number of evolutionary generations is small,but the fewer activation of nodes.(3)This paperproposes a stepwise dimension reduction modular decomposition approach for the design of multi-target circuits.And two-dimensional matrix coding scheme is used to study on the design of multi-objective circuit evolutionary.This method can effectively reduce the complexity of circuit evolution design and improve the convergence speed of circuit evolution design.