MIT's Galileo Now Aids Robots to Take Evasive Action During Disaster
FREMONT, CA: Galileo, system developed by MIT's Computer Science and Artificial Intelligence Lab (CSAIL) can now predict how real-world objects behave. The system responds similar to human analysis on the movements of an Object.
Galileo is trained itself using a combination of real-world videos and a 3D physics engine to learn the physical properties of objects and take appropriate actions. The training paves way for the model to generate a data set of objects and their various physical properties. This helps robots to predict events in disastrous situations and protect human beings.
Bullet, a 3D physics engine used to create special effects for movies and video games was initially used to feed information to Galileo. At a later stage, deep-learning algorithms were developed so that the model trains itself. The system was finally tested against human subjects to predict a series of simulations.
Katherine Noyes inks in her article for Computer World stating that researchers hopes to extend the work to more complex scenarios, with an eye on both applications in robotics as well as artificial intelligence.