Attention-based Robot Learning of Haptic Interaction

This lecture by the instructor focuses on haptic attention modeling for robot learning, where a novel architecture is presented to integrate tactile measurements over time. The robot autonomously executes haptic glances to identify object classes, achieving a high performance rate. The architecture adapts its exploration policy based on acquired data, leading to robust object class estimation. Real-world validation of the control scheme is discussed, along with enhancing haptic interaction efficiency by extending the parametrization of haptic glances.