Topics
Robot Learning
How robots acquire executable capabilities from demonstrations, data, simulation, and interaction.
Robot learning studies how robots acquire perception, prediction, planning, and control capabilities from data or experience. Common approaches include imitation learning, reinforcement learning, offline learning, self-supervised learning, and sim-to-real transfer. Learning can address tasks that are difficult to program by hand, but it does not remove the need for control, system models, or safety constraints.
This topic tracks data quality, training cost, policy generalization, evaluation design, online adaptation, and safe deployment. It distinguishes gains on a single benchmark from repeatable capabilities across environments, tasks, and robot embodiments.