Teleoperation

What it means

Teleoperation means that a person directs a robot through an interface rather than by physically moving its mechanisms. Commands can describe joint motion, a desired hand position or a larger task. The operator receives information about the robot and its surroundings, commonly through a camera view. Pollen's Reachy 2 app uses a virtual-reality headset and hand controllers to provide this connection. [1]

Remote control can happen across a room or across a network. The defining feature is the human's role in directing the activity, rather than the distance between the person and the machine. [1]

How it applies to humanoids

A humanoid interface must translate human movements into motions its own body can execute. Human and robot proportions, joint ranges and contact constraints can differ. The H2O research system trains a motion imitator using feasible, retargeted human movements before transferring it to a physical humanoid. Its operator supplies movements through an RGB camera. [3]

Teleoperation also provides demonstrations for robot learning. Reachy 2's documentation describes gathering demonstrations as one use of the VR app, and its product page distinguishes teleoperation from Python control. Recording an operator's successful attempt supplies training experience; it does not establish that the same task was completed autonomously. [1, 4]

Terminology and measurement traps

A remote operator can work alongside local controllers. An interface may request a hand pose while the robot handles the detailed motion. Read the description of human input and robot control before interpreting a demonstration as either fully manual or fully autonomous. [1, 3]

Latency has several meanings. Pollen measures the delay between a recorded controller movement and the robot reaching the corresponding position separately from the delay between a camera event and its display in the headset. A network ping is therefore insufficient to describe the operator's full experience. [2]

How to read it on humanoidversus.com

Check the robot overview and cited demonstrations for the stated control mode. A camera in sensors or a named model in ai_stack does not establish autonomous operation. Reachy 2 and NEO illustrate documented remote-control interfaces; NEO's product page describes scheduled remote expert guidance. [4, 5]

There is no teleoperation ranking here. Compare task scope, operator involvement and test conditions in the sources. When a record leaves the control mode unknown, keep that uncertainty rather than interpreting it as autonomy.

Further reading

Start with Pollen's teleoperation introduction and latency measurement notes. H2O explains motion retargeting and learned humanoid control; the Reachy 2 and NEO product pages show how manufacturers describe operator access. [1–5]