Swarm Robotics is a biologically inspired approach to the organisation and control of groups of robots. Its biological inspiration is mainly drawn from social insects, but also from herding and flocking phenomena in mammals and fish. The promise of emulating some of the efficient organisational principles of biological swarms is an alluring one. In biological systems such as colonies of ants, sophisticated cooperative behaviour emerges despite the simplicity of the individual members, and the absence of centralised control and explicit directions. Such societies are able to maintain themselves as a collective, and to accomplish coordinated actions such as those required to construct and maintain nests, to find food, and to raise their young. The central idea behind swarm robotics is to find similar ways of coordinating and controlling collections of robots.
The mechanisms that underlie social insect behaviour have inspired an approach that emphasises autonomy, emergence and distributed functioning, and avoids a reliance on centralised control and communication. This approach underlies both swarm robotics, and the closely related notion of artificial “swarm intelligence”. The term “swarm intelligence” was first coined in the context of cellular robotic systems, on the basis of the features that the simulated robotic collections shared with social insects: namely “decentralised control, lack of synchronicity, simple and (quasi) identical members” and size (Beni and Wang, 1989). Bonabeau et al (1999) describe as swarm intelligence, “any attempt to design algorithms or distributed problem-solving devices inspired by the collective behaviour of social insect colonies and other animal societies” (pg 7, Bonabeau et al, 1999). The key ingredients of swarm intelligence that they emphasise are self-organisation, and stigmergy, (indirect communication via the environment). Martinoli (2001) similarly describes the swarm intelligence approach as emphasising “parallelism, distributedness, and exploitation of direct (agent-to-agent) or indirect (via the environment) local interactions among relatively simple agents.
Swarm robotics has been described as the application of swarm intelligent principles to collective robotics (Sharkey and Sharkey 2006). The same principles of decentralised local control and communication are applied to physically instantiated robots. In swarm robotics, the emphasis is on using a number of simple robots that are autonomous, not subject to global control, and that have limited communication abilities. The reliance on local communication means that the potential problems of communication bottlenecks, or centralised failure, are avoided. The system benefits from the redundancy of using several robots: if individual robots were to fail, others could take over, and new ones could be added without the need for recalibration of communicative systems. In the same way, the activities of an ant colony need not be affected by the removal of some of its members. The simplicity of the individual robots means that they are able to respond quickly to the environment. There are also several tasks, such as exploring an environment, that can be accomplished more efficiently if a number of robots are used.
Of course, using a collection of robots creates some new problems itself (Bonabeau et al, 1999). There is the possibility of stagnation: without global knowledge, a group of robots can find themselves in a deadlock situation. Too many robots trying to reach the same location, or perform the same task could obstruct each other. Another problem is finding a solution to a task: how can situations be engineered in order that a desired solution can emerge? Nonetheless, the promise of being able to send a number of autonomous robots to perform a task, particularly in sites that are remote and inhospitable to humans, outweighs the disadvantages.