Artificial Bee Colony (ABC) Optimization

Both are swarm-based metaheuristics, but PSO updates all particles using velocity vectors influenced by personal and global bests, which can lead to faster convergence but also premature convergence. ABC uses a role-based structure — employed, onlooker, and scout bees — that naturally balances local exploitation with global exploration through its scout reinitialization mechanism, often providing better diversity maintenance on multimodal problems.

Karaboga and Basturk suggest setting the number of employed bees equal to the number of food sources (SN), with total colony size = 2×SN. A common starting point for the limit is SN × D, where D is the number of decision variables. These are heuristic defaults; systematic tuning via pilot experiments or meta-optimization is advisable for production use.

The basic ABC formulation handles only box constraints through variable bounding during initialization and scout phases. For general equality or inequality constraints, practitioners typically integrate constraint-handling techniques such as penalty functions, feasibility-based tournament selection, or repair operators. Without such additions, ABC solutions may be infeasible, and reported fitness values will be misleading.