Early work on appearance-based sensorimotor learning from demonstration. The learning algorithm used is RTAB-Map (Real-Time Appearance-Based Mapping), a real-time global loop closure detection approach for SLAM using only images. The idea is that the robot records actions (10 Hz) taught by the user between acquired images (1 Hz). After the demonstration, the robot is able to reproduce the skill learned by taking actions corresponding to different places (or sensory experiences) in the environnement.
introlab.3it.us...
introlab.github...
Негізгі бет Appearance-Based Sensorimotor Learning from Demonstration
Пікірлер