Presentation symposia highlighted problems shared by social insect colonies and human-engineered systems, and considered applications of social-insect-inspired design. Task allocation is central to the organization of work, from insect colonies to human factories. Jennifer Fewell (Arizona State University) explained how, in groups of insects and possibly humans, division of labour can self-organize when individuals vary in their thresholds for responding to task-specific stimuli. This simple mechanism can generate high levels of task specialization while allowing flexibility to respond to changes in demand. Related models have been applied to distributed control problems such as robot coordination (Krieger et al. 2000
) and flow shop scheduling (Cicirello & Smith 2004
). Craig Tovey (Georgia Institute of Technology) argued that biomimetic task allocation is most effective when the biology matches the design challenge at hand. For example, a honeybee colony's allocation of foragers among flower patches is an appropriate model for managing an Internet hosting centre because the two problems are analogous: both colonies and hosting centres must maximize resource (nectar or revenue) influx from multiple sources in a variable and unpredictable environment (Nakrani & Tovey 2007
). In general, swarm intelligence solutions are better suited to dynamic problems than to static ones, which do not present the difficulties faced by social insect colonies.
Nest-site selection is a leading model for studies of collective decision-making by social insects, and has potential for diverse biomimetic applications. When a colony fissions or its nest is damaged, it must search for and choose among new nest sites and then migrate. Consensus is built through a distributed, voting-like process: scouts independently discover, assess the quality of and recruit nest-mates to candidate sites, and the colony only commits to the best site once a quorum has been reached (Franks et al. 2002
). Nigel Franks (University of Bristol) compared the collective decision-making strategies of house-hunting ants (genus Temnothorax
) with those employed by Internet search engines, and Martin Middendorf (University of Leipzig) explained how nest-site selection can inspire algorithms for organic computing systems featuring autonomous, reconfigurable helper components. Once a honeybee colony decides on a new home, the swarm lifts off and flies up to several kilometres to the chosen nest site, even though the majority of colony members do not know its location. Kevin Passino and Kevin Schultz (Ohio State University) presented research on the mechanisms underlying swarm guidance and cohesion, and how they can be applied to distributed agreement problems in engineering, such as control of energy-efficient ‘smart lighting’ systems.
The nests of social insects, like human buildings, must accommodate and organize their inhabitants. Walter Tschinkel (Florida State University) speculated that nest architecture is shaped by natural selection to provide vital services including shelter, defence, organization of work, facilitation of movement and communication, ventilation, and microclimate control, but he conceded that because the study of ant nests has been mostly descriptive, biomimicry of nest function is probably premature. Ilaria Mazzoleni (Southern California Institute of Architecture) emphasized the adaptation of social insect nests to environmental conditions, and suggested that architects can apply similar principles to design buildings that are congruous with local climate and responsive to seasonal changes. The giant mounds built by African termites are monuments of insect architecture that have inspired passive cooling systems in human buildings. However, Scott Turner (State University of New York) discovered that Macrotermes michaelseni
mounds do not regulate nest temperature in the way previously imagined; he presented a new model for how termite mounds promote gas exchange in a process analogous to the function of a lung. Turner and engineer Rupert Soar are developing termite-inspired building materials that capture turbulent winds to manage the internal climate of buildings (Turner & Soar 2008
The coordinated behaviour of social insects can also inspire biomimetic control strategies for groups of robots, designed for jobs ranging from toxic waste clean-up to space exploration. Insect colonies are robust, scalable and function without centralized control, direct communication or a priori information about the environment—all desirable features in multirobot systems. Collaborators Stephen Pratt (Arizona State University), Spring Berman, and Vijay Kumar (University of Pennsylvania) described their use of group prey retrieval by the ant Aphaenogaster cockerelli as a model for cooperative manipulation and transport by robots. When a forager ant discovers a prey item that is too large to retrieve alone, it recruits assistance from a team of workers that lifts and carries the item back to the nest over obstacle-laden terrain. The researchers are investigating the individual actions and communication pathways that make group retrieval efficient in ants, and translating them into algorithms for controlling robot teams.
The keynote address was delivered by Eric Bonabeau (Icosystem Corporation), who uses swarm intelligence to design forecasting and optimization tools for businesses. He suggested that the general methodology of self-organization can be more instructive than specific biomimetic algorithms, and stressed two critical challenges: the ‘inverse problem’ of defining individual behaviours and interactions to shape emergence, and the exploration of a wider range of possible solutions than can be anticipated.