A special "kindergarten" for robots has been created in China, where they learn through their own trial and error rather than from ready-made human instructions. This unusual training center was founded by Tashan Technology.
The researchers' main goal is to enable robots to independently explore their environment, conduct various experiments, and learn from their own mistakes. This approach will allow the machines to continue adapting to new conditions even during operation, solving problems they were not pre-programmed to handle. For instance, if a robot bumps into a wall during training, it repeats the attempt while changing its direction of movement to avoid repeating the same mistake. Thus, every failure is converted into valuable training data.
Experts note that learning this way requires significantly more time, but it ultimate results in far more flexible and autonomous robots. In the long term, it is also expected that they will be able to share their acquired experience with one another.
This project represents an important step in the field of robotics, as it changes the traditional approach to machine learning. Previously, programmers had to anticipate all possible situations, which was nearly impossible given the unpredictable conditions of the real world.
Although the model of learning independently from mistakes requires more time, it creates systems capable of adapting to unexpected changes. In the next stages, it will be remarkable to see how effective the exchange of experience between robots will be.
