Language-Conditioned Robot Control
Overview
The field studies how robots can turn spoken or written instructions into actions that fit what they perceive and what their bodies can do, from object manipulation and navigation to humanoid movement. Approaches include end-to-end vision-language-action policies, language-model planners paired with low-level controllers, and reusable skills or robot programs; training draws on demonstrations, simulation, reinforcement learning, and motion data. Work is expanding from simulated benchmarks such as LIBERO to real robots, including mobile platforms, drones, arms, and humanoids, with added tools for user feedback, safety checks, and adaptation. A central challenge is making language understanding reliable under ambiguity and distribution shift while ensuring that generated actions remain physically feasible, safe, and responsive in closed-loop execution.
Latest papers
74 papers12 papers in the last four weeks, up 71% on the four weeks before. 0.1% of all new papers.