Abstract :
Glowworm Swarm Optimization algorithm is applied for the simultaneous capture of multiple optima of multimodal functions. The algorithm uses an ensemble of agents, which scan the search space and exchange information concerning a fitness of their current position. The fitness is represented by a level of a luminescent quantity called luciferin. An agent moves in direction of randomly chosen neighbour, which broadcasts higher value of the luciferin. Unfortunately, in the absence of neighbours, the agent does not move at all. This is an unwelcome feature, because it diminishes the performance of the algorithm. Additionally, in the case of parallel processing, this feature can lead to unbalanced loads. This paper presents simple modifications of the original algorithm, which improve performance of the algorithm by limiting situations, in which the agent cannot move. The paper provides results of comparison of an original and modified algorithms calculated for several multimodal test functions.
