Brain–Computer Interface for Humanoid Robot Control Adaptation

Modern developments in both robotics and neuroscience have made it possible to show off the first brain–computer interfaces (BCIs) for commanding robots with human-like characteristics. But earlier BCIs depended on command and control at a higher level, depending on behaviours that were hardwired in. Conversely, the BCI user may have a heavy cognitive strain due to the monotony of low-level control. To overcome these issues, an adaptive hierarchical method of brain– computer interaction was suggested. In this method, users train the BCI system to do new tasks automatically; thereafter, they may call upon these abilities directly as high-level instructions, which eliminates the need for repetitive manual control. This research delves into the use of hierarchical BCIs for training and controlling a PR2 humanoid robot. Consider using explicit command sequences to enable users to design complicated tasks with numerous state spaces. Three people successfully trained and controlled the PR2 robot with humanoids using a hierarchical EEG-based BCI to pour milk over cereal in a simulated domestic scenario. The first hierarchical BCI training for non-navigational tasks is presented. The example is the first to use one model for training a difficult problem with several state spaces.

Перевод пока недоступен