A scout bee is the exploration agent in an artificial bee colony (ABC) algorithm. Each food source represents a candidate solution. While employed bees improve nearby candidates and onlooker bees favor promising ones, scouts search for new candidates in other parts of the problem space. This division balances local improvement with the chance of finding something better elsewhere. section 10.5
The role comes from honeybee foraging: foragers gather from known sources, while a smaller number of scouts look for new ones, especially as existing sources deplete. An artificial employed bee can likewise switch to scouting when its assigned solution has not improved after a set number of attempts. It abandons that source and randomly searches for a replacement. This helps the search escape local optima, but a newly discovered candidate is not guaranteed to be better. section 10.4 section 10.5
In an ABC iteration, scouting follows two phases focused on exploiting known solutions:
- Employed bees search near their current food sources for improvements. section 10.5
- Onlooker bees select sources according to fitness and search around them. section 10.5
- Scouts replace abandoned sources with newly discovered ones; the best solution found so far is retained. section 10.5
Algorithm 10.1 Artificial bee colony algorithm
Initialization Phase: population of candidate solutions (food sources) are initialized REPEAT Forager Bee Phase: Each forager bee goes to a food source in her memory and determines a closest source, then evaluates its nectar amount and dances in the hive Onlooker Bee Phase: Each onlooker bee watches the dance of forager bees and chooses one of their sources depending on the dances, and then goes to that source. After choosing a neighbor around that, she evaluates its nectar amount. Scout Bee Phase: Abandoned food sources are determined and are replaced with the new food sources discovered by scout bees. Memorize the best food source (solution) achieved so far. UNTIL (termination criteria are met)
The algorithm outline shows where abandoned sources are replaced and how the scout phase fits between local search and best-solution tracking. section 10.5
Trying scouts on a function
For Rosenbrock-function minimization, a Python ABC example uses two dimensions, 50 bees, 1,000 iterations, and a 100-trial limit. Its reported run found a candidate near (1, 1). Adjusting the trial limit changes when stalled sources are abandoned; compare runs rather than treating one stochastic result as guaranteed. A separate `mealpy` example runs `OriginalABC` with a population of 50 and a limit of 15. section 10.5 section 10.6
