In any robotic arm or a robotic mechanism that has more than 3 degrees of freedom, a huge part of the programming goes to program the movement themselves. How about building a robot, regardless of how the motors and joints are connected? And what if a robot can be built with no knowledge of itself, the robot becomes aware of the way it is physically built?
Columbia Engineering researchers have actually made it possible. They create a robot arm that learns how it is connected, with zero prior knowledge of physics, geometry, or motor dynamics.
Initially, the robot has no idea about its shape, its motors, and its movements. After one day of trying out its own outputs in a pretty much random fashion and getting feedback on its actions, the robot creates an accurate internal self-simulation of itself using deep-learning techniques.
For the study, Lipson and his Ph.D. student Robert Kwiatkowski used a four-degree-of-freedom articulated robotic arm. The first self-models were inaccurate as the robot did not know how its joints were connected. After about 35 hours of training, the self-model became consistent with the physical robot to within four centimeters.
After that, the self-model performed a pick and place task which allows the robot to recalibrate its original position between each step along the trajectory.
“This is perhaps what a newborn child does in its crib, as it learns what it is,” said Hod Lipson, a professor of mechanical engineering who worked on the robot, in a press release. “We conjecture that this advantage may have also been the evolutionary origin of self-awareness in humans. While our robot’s ability to imagine itself is still crude compared to humans, we believe that this ability is on the path to machine self-awareness.”
Furthermore, for testing whether the self-model could detect damage to itself or not, the researchers 3D-printed a deformed part to simulate damage. And the best part is the robot was able to detect the change and re-train its self-model. That new self-model enabled the robot to resume its pick-and-place tasks with little loss of performance.
As the internal representation is not static, this helps the robot to improve its performance over time. Besides, it also allows it to adapt to damage and changes in its own structure. This could help robots to continue to function more reliably when there its part starts to wear off or, for example, when replacement parts are not exactly the same format or shape.
Yet the robot is far away from writing a poem or a note. But, Lipson hopes that the project as an important step toward understanding how humans learn to conceive of themselves. And one day they will definitely be able to build robots that understand themselves as we do.
