A more realistic genetic algorithm

Genetic Algorithms (GAs) are loosely based on the concept of the natural cycle of reproduction with selective pressures favouring the individuals which are best suited to their environment (i.e. fitness function). However, there are many features of natural reproduction which are not replicated in GAs, such as population members taking some time to reach puberty. This thesis describes a programme of research which set out to investigate what would be the impact on the performance of a GA of introducing additional features which more closely replicate real life processes. The motivation for the work was curiosity. The approach has been tested using various standard test functions. The results are interesting and show that when compared with a Canonical GA, introducing various features such as the need to reach puberty before reproduction can occur and risk of illness can enhance the effectiveness of GAs in terms of the overall effort needed to find a solution. As the method simulating the nature rules, Cardiff Genetic Algorithm (CGA) introduces several features to each individual in programming modelling the real world. Each individual of the population is given a life-span and an age, the population size is allowed to vary; and rather than generations, the concept of time steps is introduced with each individual living for a number of time steps. An additional feature is also discussed involving multiple populations which have to compete for a limited resource which can be thought of as “water”. This together with an illness parameter and accidental death are used to study the behaviour of these population