This ad will close in 15 seconds

From : albertaBCP
Added: May 1, 2009
A LEGO NXT robot that is capable of classic navigational tasks without any 'thinking' involved. A common experiment used to explore how animals navigate is the 'reorientation task', which involves seeing how animals behave in arenas that vary slightly from ones they have been trained in. Curiously, the same mistakes are made in all animals tested, ranging from goldfish to human infants. For twenty years, the prevalent theory to explain these mistakes involves the assumption that the brain generates some mental 'map' of its surroundings. This is of great interest to modern robotics engineers and artificial intelligence researchers, since finding your way through a new environment is a difficult task for most robots. One current technique for robotic navigation is SLAM - Simultaneous Localization and Mapping, which essentially involves building a representative 'map' and navigating within it at the same time. This is, naturally, very processor-intensive. By a stroke of luck, when exploring the behaviour of a simple LEGO NXT robot design that cannot think (it can merely sense and act), we stumbled upon a robot that happened to perform the reorientation task with exactly the same errors that animals make. The machine isn't told anything about its environment (it doesn't know what a corner is, let alone how many there are in the arena, for instance), nor does it plan routes: It merely sets each motor's speed to a simple function of the sensor input on the other side of the robot. This robot demonstrates that it is possible that navigation in animals is a simple process of interaction between the agent and its environment without any mental maps at all -- but testing this claim is where Real Science (with LEGO!) comes in. Since the robot can navigate complex environments (even ones that shift about quickly; it isn't shown in the video, but we've released it in large crowds and it gets through them just fine, only to get lost in distant hallways and spook the janitor) without any "mental map", this robot is intended to rebut the necessity of SLAM. This explains its name: antiSLAM. The robot design started out as a Braitenberg Vehicle 2 (see our earlier videos) that used a different type of sensor. It finished as essentially two Vehicle 2s of different strengths that share the same set of motors, to give you a sense of how complex it is (i.e. not very). LEGO fans will notice that we used two ultrasonic sensors on the same robot, which is usually considered taboo in Mindstorms circles. We say, "why not?" -- and happened to get a really interesting machine out of it. This is the version of the machine that appears in our upcoming book, From Bricks to Brains: The Embodied Cognitive Science of LEGO Robots. Building instructions and program code will be available in the book when it is published (soon). Arena images are based on originals from Wystrach and Beugnon 2009, a research paper on the reorientation task in ants (which, unusually, reported on the paths the animals took). Music (of the YouTube version of this video, due to copyright concerns) is ██████ ██ ██████, by █████ ████.
Category : Tech

Video Categories:

Also Try: