FOR IMMEDIATE RELEASE
CMU UNVEILS NEXT-GENERATION AUTONOMOUS SYSTEM
The Advanced Manipulators Lab at Carnegie Mellon University has announced the launch of a next generation, fully autonomous system for field trials, that project scientists predict will set a new standard for overall systems performance in robotics. The neural network-based system, dubbed "Mary-Kate" by its creators, is equipped for autonomous navigation in rugged terrain with facilities for cooperative and dextrous manipulation. Locomotion is achieved not with wheels, but with an advanced set of four manipulators that are dynamically reconfigurable to provide four-legged gaits -- or, researchers hope, to eventually provide dynamic stability with a two-legged gait after several months of training.
"Mary-Kate" is equipped with a vast array of sensors and real-time processing hardware including stereo vision, tactile arrays, thermal sensors and an advanced passive sonar system that spatially locates objects in the environment by the sounds they emit. The passive sonar relies on a stereo pair of sensitive sonic receivers that time-correlates the input signals to judge angular position by triangulation. The neural network computer, which literally contains "billions and billions" of hidden units, uses the amplitude of the sonic signals and a modifed table-lookup approach to estimate the distance to known objects whose nominal signals appear in the database.
The most astonishing aspect of the announcement is the diminutive size of the overall system. Weighing it at a mere 8 pounds, 11 ounces and measuring only 22.5 inches in length, Mary-Kate packs incredible capabilities into a very small package.
Principal Investigators, Richard Meredith Voyles and Kathleen Marie Voyles, say Mary-Kate, which is short for Meredith Kathleen, represents a return to a recently abandoned trend within the Advanced Manipulators Lab that prefers a feminine naming convention for developmental systems.
"We `severed the umbilical' at 9:57 this morning and the system has operated fully autonomously ever since," said PhD-hopeful Voyles. "I expect to learn alot from this experiment."