Swarm Intelligence

WE can learn much about decision-making from nature. The honey bee swarm is considered to be one of the best examples of collective decision-making outside the human domain.
In nature, strong honey bee colonies reproduce by dividing themselves. The mother queen and about two-thirds of her workers leave the nest, fly to a nearby support such as a branch or hedge, and form a swarm cluster. Scouts then fly from the cluster to search the surroundings for potential new nest sites. Scout bees must evaluate various potential nest sites and decide on a final site before the colony runs out of food.
For the bees, an effective decision-making strategy is needed to cope with all the informational variations such as numbers, quality and location of alternative nest sites, and the order of site discoveries. The initial information gathering phase gradually gives way to the decision-making phase, and it is done in a virtuous cycle of decision-making by the scout bees when better sites are discovered and poorer sites abandoned.
The scouts supporting the better quality sites fly more trips and garner more scout bees to join them. The total scout numbers present at each site is a good measure of the total evidence in favour of a particular site as the swarm’s new home. This is carried out until a critical quorum threshold is reached at one of the sites which becomes the preferred nest site.
The scout bees are interacting locally with one another and with the environment. They follow very simple rules without any centralised control structure dictating how they should behave. Yet, such interactions lead to emergent “swarm intelligent” behaviour unknown to the individual bees. Swarm Intelligence (SI) was first introduced in 1989 by Gerardo Beni and Jing Wang to define the collective behaviour of decentralised, natural or artificial self-organised systems. Besides bee colonies, other examples of SI in natural systems include ant colonies, bird flocking, animal herding, microbial growth, and fish schooling.
By observing the flocks, herds and colonies of animals, we learn different kinds of dynamic group behaviours where a large number of individuals without supervision can accomplish difficult tasks by following simple rules when they interact with each other. They distribute problem solving among many individuals, allocate resources efficiently and also adapt rapidly to changes in the environment. Instinctively, they self-organise in a clever manner.
This has inspired scientists to capture their behaviour in algorithms, which have useful applications for organisations to optimise their complex business operations, to tap into wisdom of crowds, and to pool information in order to improve on one another’s insights. Two lessons may be drawn from the smart swarms as suggested by Peter Miller, author of Smart Swarm (using animal behaviour to change our world). Firstly, when we are working together, we can lessen the impact of uncertainty, complexity and change.
The swarms have relied on local knowledge of diverse information, simple rules that have no need for complicated computation skills, repeated interactions to amplify key signals for speedy decision making, quorum threshold to arrive at a quality decision, and randomness in individual behaviour.
The last principle helps to overcome negative aspects of social proofing so as to avoid group think and be stuck in a rut. By applying the above principles, businesses can source much wisdom from the individual stakeholders.
Secondly, group members need not surrender their individuality. This may augur a conducive working space where we add values to a team or organisation by bringing something authentic and original to the table without blindly following the herd.
At times this means standing up for what you believe in, and canvassing support for your idea. This is demonstrated by the scout bees vying for the attention of other bees to join them to their preferred nest sites.
Interestingly, a sampling of past researches shows the following: firefly algorithm for solving non-linear programming problems, strategy adaptation based bacterial foraging algorithm for numerical optimisation, ant colony optimisation (ACO) algorithm for deterministic optimisation, swarm robotics system, particle swarm optimisation for production line efficiency, and swarm behavioural inversion for undirected underwater search.
The last example is a potentially useful algorithm for the trans-oceanic search of the missing planes and ships. SI-based techniques are also used for controlling unmanned military vehicles, and perhaps in the near future, the autonomous (driverless) vehicles on our streets.
Contact Director of VU Postgraduate programmes, Dr Hendry Ng at hendryng@sunway.edu.my for more details about Victoria University Master of Business Admintration (MBA) and Master of Business (ERP Systems).