Genetic algorithms (GAs) are population-based optimisation tools inspired by evolution and natural selection. They are applied in many areas of engineering and industry, on increasingly complex problems. To improve the performance, the new algorithms have a tendency to be derived from sophisticated mathematical and computational mechanisms, where many biological and evolutionary advances have been neglected. One such mechanism is multi-level selection theory which has been proposed as being necessary for evolution. Previously, an algorithm developed using this theory as its inspiration has shown promising performance on simple test problems. It proposes the addition of a collective reproduction mechanism alongside the standard individual one. Here the algorithm, Multi-Level Selection Genetic Algorithm (MLSGA), is benchmarked on more sophisticated test instances from CEC '09 and compared to the final rankings. In this instance a simple genetic algorithm is used at the individual level. The developed algorithm cannot compete with top algorithms on complex unconstrained problems, however it shows interesting results and behaviour, and better performance on constrained test functions. The approach provides promise for further investigation, especially in integrating state-of-the-art individual reproduction methods to improve the performance and improving the novel collective mechanism.
