Page 21

CIM29

N°29 I quarterly I APR IL 2013 Focus | 21 w A New Breed of Robots Angels, a composite swimming robot that ripples through the water like an eel, should soon be making waves in the pool of the IRCCYN laboratory.1 Magnetically attached in a single file, its eight propeller-driven 25 cm-long modules are able to split up, swim by themselves, and communicate to reassemble. How is it possible? “The modules generate electric fields to probe their surrounding environment and navigate. This method, called  electrolocation, is used in nature by elephant fish and other animals that live in murky waters,” explains Frédéric Boyer, the IRCCYN roboticist who heads the research project.2 Angels is funded by the European Union and scheduled for completion, with final system trials, in September 2013. 01 Diagram of Angels, the eel-like robot, next to its model. 02 Angels’ component modules were recently assembled for a final test to be conducted in September 2013. AUTO NOMOUS MACHINES Why does it swim like an eel? “It affords better maneuverability for a robot with this kind of elongated shape,” adds Boyer, “and the detachable modules make it possible to survey a large area relatively quickly. Each module can swim in a confined zone, which could have applications in military surveillance, industrial endoscopy—or even medical endoscopy with ultra-miniaturization.” Electrolocation technology gives the man-made creature an exceptional degree of autonomy, even in low- visibility situations. For robotics experts, finding a system that lets machines move through space independently—without remote control or pre-programmed routes—has been a long-standing dream. It offers the possibility of sending robots into unknown or humanly inaccessible environments, or assisting the elderly in complex daily tasks. “To meet this challenge, most researchers tried to mimic human vision, with artificial intelligence programs to process the images,” Boyer explains, “but this model proved too ambitious.” In terms of navigation, emulating human vision has indeed led to many robots with very complex digital systems able to model the geometrical space around them in the form of cubes or cylinders. Yet today, most are still unable to make their way through a maze. Meanwhile, a simple ant has no trouble navigating with only a few hundred thousand neurons, compared to the human brain’s one hundred billion. Notwithstanding its limited powers of perception, a tiny insect can negotiate unknown territory better than any high-IQ robot ever developed. A CHANGE OF TA CK Today’s biorobotics researchers are trying a different approach. “We no longer consider intelligence to be the sole product of a brain structured like a calculator, as envisioned by Alan Turing,3” Boyer explains. Programming a chess champion is not the be-all and end-all. “This type of intelligence is far removed from the environment, like a brain in a jar that could only be attached to our robots,” the researcher emphasizes. “Instead of developing built-in intelligence, we are now trying to create embodied intelligence, which is closely tied to body and morphology, and found in so-called ‘primitive’ animals.” In short, an intelligence that is not necessarily highly developed, but linked to the body’s sensory interactions with its environment. This is the case of the Angels robot, whose casing is covered with 13 types of novel sensors that have been enhanced for future applications involving submarine navigation in dark or muddy waters. The sensors are capable of generating an electric field and detecting any disturbance in the vicinity. “The body is both the locomotive organ and the sensory surface through which it perceives its surroundings,” Boyer says. “This makes it possible to establish direct links between sensations and actions, without going through an abstract representation of the entire space, which uses up a lot of computing power.” These high-speed, high-quantity links, called sensorimotor loops, take the form of algorithms defined by computer scientists in close collaboration with neurobiologists. “With this sensorimotor approach, the electric field interacts with the space around it, triggering immediate, appropriate reactions in the robot,” adds Boyer. ENTO MOLOGICAL INSPIRAT ION Mimicking animals, especially insects, offers many other advantages like unprecedented agility, compactness, and 01 El ectrolocation. The ability to sense surroundings by capturing the reflections of an electric field on the surface of obstacles, preys, etc. 02 © N. Servagent/École des Mines de Nantes © picture partners/fotolia © c. frésill on/CNRS photothèque


CIM29
To see the actual publication please follow the link above