By Lina Sorg
Swarm robotics is a fairly new technique that coordinates systems comprised of many of low-cost, autonomous robots. The field involves the study of both the physicality and displayed behaviors of robots, and researchers hope that the interactions of these robots with each other and with their environment will yield beneficial collective behavior applicable to varied domains and timescales. For example, swarm robots find potential application in disaster response, environmental monitoring, collective transport, and mapping. The concept of artificial swarm intelligence draws inspiration from biological collectives, specifically the study of biochemical systems, social insects, and other creatures that exhibit swarm behavior (swarm intelligence).
In a minisymposium talk entitled “Swarm Robotic Control Strategies Inspired by Biological Collective Behaviors” at the SIAM Conference on the Life Sciences, Spring Berman (Arizona State University) addressed the complications and opportunities associated with robotic swarms. Individual swarm robots should be kept as simple as possible; this ensures that they act productively and meaningfully at the swarm level, rather than the individual level. Swarm robots are also subject to many constraints. As in natural swarms (i.e. in the insect world), the robots are only able to access local information they discover during exploration. They have no global localization, no prior data about the environment, and no local or inter-robot communication. Conversely, the robots do have traits analogous to biological collectives, including local sensing and limited computational abilities. Each individual robot is not told exactly what to do, but rather programmed to operate in a specified parameter. Possible programmable tasks include mapping and covering boundaries, mapping and covering interiors of regions, and returning to the target destination while avoiding obstacles.
Because robots move around randomly, they can be modeled as chemically-reacting molecules. Berman and her team use a “top-down” controller design with stochastic behaviors, and employ partial different equations to model spatial dependencies. With influences from fluid dynamics and chemical kinetics, Berman describes task transitions, robots’ roles, at the level of both the individual and the population; probability groups, for example, can help program directive motion. A potential application is commercial pollination via robotic bees. The aforementioned modeling techniques allow the robot to switch between flying and hovering states, and optimization can help robots specify distribution of a flower’s pollen. Targeting drug delivery to tumors is yet another application, where reaction diffusion models enable researchers to design a strategy for particles to deeply penetrate the tumor.
Berman hopes to continue building on her work with more advanced technologies. Numerous undertakings can benefit from the implementation of swarm robots, including distributed sensing, mining, and agricultural foraging. However, the most promising use of swarm robots is likely in disaster rescue missions, during which swarms could use infra-red sensors to detect living victims in areas too unsafe for rescue workers. More controversial is the potential use of swarms in military combat.