At the moment, no matter how capable nor was the robotic mechanism, it in most cases operates according to pre-defined algorithms (or not controlled) and cannot foresee the consequences of their actions. But scientists from the University of California at Berkeley has created a new learning technology machines, which enables them to predict the consequences of a particular act, even when interacting with objects that previously the robot is not met.
About the new development the newspaper writes EurekAlert. According to their reports, the technology is arranged as follows: in the first stage the robot without operator intervention using images obtained from cameras, independently moving objects, on the basis of which is built a predictive model of behavior. After several attempts she “remembered” and used in the future when meeting with a new or similar seen objects. The camera of the robot to calculate the possible sequence of actions, but so far, however, only a few seconds ahead. Although this is already enough in order for the robot decision-making in the new conditions without human intervention. As stated by one of the researchers, Frederick Ebert,
“People all my life learning to communicate with the outside world practically unaided by trial and error, earning need experience and acquiring skills. We managed to build a robotic system that will use a large number of Autonomous data collected to acquire useful skills and knowledge acquired during training, the machine can use in the future when, for example, the movement of objects”.
Robots taught to predict the future
Vladimir Kuznetsov