Recently, engineers at Columbia University have created a robotic arm that they can imagine, which marks a step forward in the development of self-aware robots. The ability to have self-awareness and imagination in different situations is unique to human beings and is a key factor in our human progress. The ability to imagine the future helps humans move toward this goal, and the end of the analysis of past experience, which some animals can do, can also help us to learn to avoid mistakes.
Moreover, human self-imagination is not static. As time goes by, we will adapt and change with the changes of environment and experience. But this is not the general learning style of robots. Robots generally acquire experience and knowledge through manual design simulation and modeling, or trial and error.
self conscious? This robot can imagine what his arms look like.
Engineers at Columbia University built a robotic arm that didn't know much about physics, geometry, and motor dynamics, but after a day of intensive computing, it had a sense of self-simulation, and then the arm used this self-simulation To adapt to different situations and tasks. Hod Lipson, professor of mechanical engineering and director of the Creative Machines Laboratory, said, "If we want robots to become independent and quickly adapt to unforeseen scenarios, it is important to make them self-aware."
Hod Lipson and his Ph.D. student Robert Kwiatkowski created a very simple robot with four free articulated arms. When the robot first started, its arm was basically moving irregularly. Then, it created the first self-built model using deep machine learning, and after 35 hours of training, the self-built model was greatly improved, consistent with the robot's body, with an error of no more than 4 cm.
Then they started experimenting. They asked the robot to pick up things and put them back in place. The robots performed well. The robots can recalibrate the motion trajectory according to the internal self-built model. The robot arm can accurately grasp the objects on the ground. And put them 100% into the trash. In a closed-loop control system, the success rate of the robot arm is 100%. In the open-loop system of the internal self-built model, the success rate of the robot arm reached 44%.
Robert Kwiatkowski said, "It's like holding a glass of water with your eyes closed, even for humans." In order to change the way the robot operates, the engineers put in a part, and the robot can re-adjust its self-built model and pick up the object again.
Hod Lipson said, “This is more like a newborn doing something in a crib. We speculate that this advantage may also be the origin of the evolution of human self-consciousness. Although the robot’s self-awareness is simple compared to humans, We also believe that the development of self-awareness can greatly advance the progress of robots."