A population minimisation process for genetic algorithms and its application to controller optimisation

This paper suggests a process which helps reduce the execution time for genetic algorithms by removing the redundancy associated with the saturation effect found in the later generations. The process considered minimises the population size as similar individuals occur in the fitter members of the population. As the population size reduces the number of crossover operations decreases and the apparent mutation rate increases. This increase in variation allows better avoidance of local optimal solutions. The process is evaluated by considering results obtained from its application to a submarine controller optimisation problem.