A group of researchers from the Massachusetts Institute of Technology presented a training system for robots called C-LEARN. The new development allows robots not only to acquire new skills, but also to transfer the accumulated experience to other robots, which makes the learning process faster and more efficient. In addition, thanks to the C-LEARN "set vector" training can even people who do not know anything about programming.
Experts responsible for the development of the Laboratory of Informatics and Artificial Intelligence (Computer Science and Artificial Intelligence Laboratory, abbreviated to CSAIL). To begin the learning process, a person needs to "show" the robot how to do this or that job, and set the basic rules that the robot should follow. The rest will be done by the robot on its own, after which it will be ready to transfer the acquired experience.
The C-LEARN system is quite simple. It is to provide information about the ultimate goal and the constraints that will be in place as it progresses. Suppose, if the robot has to install a steering wheel and a wheel in the car, the operator should specify not only the sequence of actions (what and where to screw), but also to set a number of parameters by which the robot will operate. That is, he will have to distinguish between the purpose of the object (since they are both round in shape), fasteners and so on. Then the robot performs the task, and the person either approves the result and puts it in the database, or makes the necessary adjustments. In addition, the system includes two types of training: the above-described training on examples and training using algorithms of high-level programming.
Once the robot has acquired the necessary skills, they can be transferred to another machine. In a series of experiments it was found that in the case of training a robot man the learning process was successful in 100% of cases, but when the robot trained another robot, the success rate was only 87.5%. However, during the tests, CSAIL employees trained Optimus to neutralize a certain type of explosive, after which Optimus transferred this experience to another robot called Atlas and both robots used this skill with equal efficiency.