The development of a robot that can learn to move around the world in a completely autonomous way, for many members of the scientific community, is only possible in the universe of fiction. However, a group of researchers at Google Robotics managed to use the power of artificial intelligence to help a small robot learn to walk without human intervention within two hours.
Instead of training the robot's algorithms in a virtual environment before moving on to reality, the team of researchers opted for a different method. Since the simulation of certain surfaces is a complex and time-consuming process, the robot was tested directly in a real environment.
Experts say in a recently published study that the developed algorithm is able to learn more effectively. When walking on surfaces with great variations in texture, the robot was also able to adapt more quickly to environments with inclines, steps and obstacles. Upon reaching, for example, the edge of a certain surface, the automaton recognizes its position and starts to walk backwards alone.
Through many trials and errors, the robot eventually learned to take its first steps. However, the process had to rely on constant human vigilance and, at an early stage, some intervention by the researchers.
In addition to trying to apply its discovery to other robots, the Google Robotics team will now try to overcome one of the limitations of the system created. The robot uses motion capture technology to help you find your location. However, experts indicate that it is not possible to replicate the system in automats in the real world.