The demonstrations focused on how MCS systems grasp objects, adapt to obstacles, change speed and other capability enhancements MCS researchers put into the robotics, DARPA said Wednesday.
Mimicking child cognition’s core domains for objects, agents and places is what MCS looks to achieve in its computational model development initiative. The program researchers employed simulated training to support their efforts.
“These experiments are important milestones that get us closer to building and fielding robust robotic systems with generalized movement capabilities,” said Howard Shrobe, MCS program manager in DARPA’s Information Innovation Office.
As part of the experiments, a rapid motor adaption algorithm was created by University of California, Berkeley researchers to enable four-legged robots to adapt to and navigate through changing terrain.
Meanwhile, a bipedal robot was tested by Oregon State researchers to demonstrate its ability to use proprioceptive feedback to learn how to carry dynamic loads.
Other algorithms were developed and tested in line with the MCS program. They were trained entirely in simulation.
“By focusing on commonsense, we are creating the possibility for systems to have the flexibility of human learning and the breadth of human knowledge,” Shrobe said.