Study On Some Issues Of The Genetic Algorithms Under The Influence Of The Noise

In recent years, the different kinds of problems which are on the basis of the optimization like robot path planning, production plan and resource scheduling, and discrete events dynamic system monitoring, have a complex environment to solve.Uncertain environmental factors, the mistakes of the training pattern and human factors can all cause the problems that are under the influence of noise. So, when the traditional optimization methods for the global optimal solution are implemented and actual situation deviates from the hypothesis, these methods can’t maintain optimal,but also lead to optimize quality decline seriously. In this paper, the genetic algorithm(GA) was studied under the noise interference and did some surveys from following aspects:(1) From the basic knowledge of genetic algorithm, the introduction of noise variables, genetic algorithm are studied under noise environment. Because the noise environment can lead to problems of optimal offset, This paper use the average effective objective function method for noise reduction processing, try to make the influence of noise on algorithm optimization to a minimum.(2) Problems of noise jamming decision variables were studied. Using resample to obtain the average effective objective function to reduce noise interference. This paper introduces the related concepts of genetic algorithm robustness and brings in variance as robust constraint conditions of genetic algorithm. Robust optimization are implemented under the constraint condition, from which the solution comes is more stable and anti-interference performance is stronger under this method.(3) This paper studies the noise interference fitness function evaluation problem.The method of resample are still used to obtain the average effective objective function to reduce noise interference. On the basis of this method, this paper put forward an improved genetic algorithm that can be used to reduce the influence of fitness evaluation in a noise environment. Using the global zone convergence rate and the global convergence accuracy as a new index to evaluate genetic algorithm performance under noise interference fitness function. Experimental results show that the improved genetic algorithm can improve the optimization results.