Improved Genetic Programming Research Based On Evolvable Hardware

The evolvable hardware is a new design methodology that integrates theevolutionary algorithm into the Programmable Logic Device (PLD). This paper recallsthe platforms of programmable devices and the background knowledge of evolvablehardware including basic principles and classifications of evolvable hardware. Inaddition, four types of evolutionary algorithm are illustrated and compared in detail,including their own main ideas and characteristics.Due to the automatic designing methodology of the evolvable hardware circuit, animproved genetic algorithm is applied from the encoding point of view. Firstly, theinitial population generation is realized by the usage of the group crossover operator,which leads to an enhancement of the quality of the initial population。Secondly, inorder to increase the diversity and superiority of the generated offspring, a new methodwhich combines the optimization and competition methods is applied for the choice ofoffspring. Owing to the disadvantage of the crossover operator in destroying thestructure of algorithm tree blocks, an improved multi-crossover operator by consideringthe method of taking the optimum in terms of their fitness is developed. Finally, toguarantee the diversity and superiority of individuals as well as to control the number ofgenerated nodes, a dynamic nodes algorithm is proposed to control the complexity.The contribution and the symbols regression problems in this paper are simulatedon the simulation experiment platform Matlab. Then, a design of simulation experimentof the full adder circuit is realized on the software environment Quartus9.0and thesimulation platform NiosⅡ9.0IDE by constructing a NiosⅡsoft core. The simulationresultants show that the proposed algorithm improves the quality of the initialpopulation and the complexity of individuals is well-controlled during the evaluationprocess. The average number of convergence is increased, and the mean amount ofiteration is decreased evidently.