Teleoperation allows a human operator to control a robot from a distance, seeing through its cameras and commanding its movements via input devices such as joysticks, exoskeletons, or VR hand controllers. In humanoid robotics, teleoperation serves a dual purpose: it enables remote task execution in hazardous or inaccessible environments, and it provides a powerful mechanism for collecting the demonstration data that fuels imitation learning systems.
Several companies have made teleoperation central to their strategies. Sanctuary AI uses teleoperation extensively with its Phoenix robots, employing human operators to perform warehouse tasks while simultaneously training autonomous AI systems from the collected data. Figure and 1X Technologies have similarly invested in teleoperation rigs that capture high-quality kinematic and sensor data from human demonstrations. The quality of this data directly impacts the effectiveness of the learned autonomous policies.
The technology is advancing rapidly. Low-latency 5G and satellite communication enable teleoperation over long distances, while haptic feedback devices give operators a sense of what the robot is touching. Apple Vision Pro and Meta Quest headsets are being repurposed as affordable teleoperation interfaces. As autonomy improves, the industry expects a gradual transition from full teleoperation to shared autonomy, where the robot handles routine subtasks independently and calls for human guidance only when encountering novel situations. For deeper coverage, see HumanoidIntel.