Booster T1 teleoperation: building the VR stack Booster does not ship, and moving to the T2

The Booster T1 is a RoboCup robot first: a compact biped with an open SDK, an Isaac Gym locomotion framework and an official RoboCup demo that plays full matches, and Booster now headlines RoboCup 2026 championship wins on its home page. It is also a robot whose maker never shipped a teleoperation stack. Press coverage called the T1 a platform for teleoperation-first work, but the code Booster publishes is for locomotion, deployment and robot soccer, not for driving the arms from a headset. In 2026 Booster replaced it on its website with the T2, a larger 31-DoF robot with optional hands and Jetson Thor compute. This page covers both: how to teleoperate a T1 you already own, and what changes when you move to a T2.

T1 or T2: where things stand in October 2026

For teleoperation the important jump is the arms. Seven joints per arm and a 3-DoF waist let a T2 present a hand at the position and angle an operator's hand is at, which is what makes hand-tracked VR data usable. The T2 product page lists secondary development as supported on both editions, so you do not pay extra for SDK access.

As of October 2026, foreign-produced advanced robotic devices are on the FCC Covered List under FCC public notice DA 26-786 (July 28, 2026), unless the Department of War grants them a Conditional Approval. The FCC definition covers mobile robots, including humanoids, quadrupeds and autonomous mobile robots, that weigh over 4.4 lb with any base or dock and have a sensor, network connectivity and control software, when they do not qualify as US-made under the federal Buy American test, whatever the nationality of the maker. Covered models cannot receive new FCC equipment authorizations, which most devices need before they can be imported or sold in the US, while the FCC says models it authorized earlier can still be imported, sold and used; US buyers should confirm the current status with the seller.

What Booster gives you to build on

Building a teleop stack on a Booster robot

With no official teleop app, every Booster data program assembles the same pieces. The pattern that works keeps balance on the robot and puts only intent on the network.

Data from a custom Booster stack

Booster's public repositories include robot descriptions and example motion data, but no teleoperation datasets, so there is no default schema to copy. That is a freedom and a risk. Decide the shape of an episode before the first session and write it down in a dataset card that ships with the data.

Recording straight to LeRobot format keeps the data usable with the open training tools without a conversion step. Our data collection guide covers formats and quality checks.

Hands or grippers on the T2

The T2 offers a standard end effector, a gripper or a dexterous hand, and the choice shows up in the price. On US storefronts in August 2026, Grabarobot found the Edu and Pro at $42,999 and $48,999 with a gripper, and $48,999 and $54,999 with a dexterous hand. For a first dataset, the gripper is usually the better buy: a pinch retargets cleanly from hand tracking, operators reach a consistent technique quickly, and the clean-episode rate is higher. A dexterous hand is worth it when the task needs finger-level work, such as tool use or in-hand reorientation, and then operators need more practice before their episodes count.

Whole-body or upper-body data

Booster positions the T2 for whole-body tasks: legs, waist, arms, head and end effectors coordinated for bimanual manipulation and manipulation while walking. A joystick-plus-VR setup records upper-body data with a walking command, which covers most tabletop and shelf work. True whole-body data, where the legs are part of the demonstration, needs body tracking for the operator (a motion-capture suit or a VR full-body solution) and a whole-body controller on the robot. With Jetson Thor on the Pro, the compute for that controller is on board; the integration is yours to build.

Latency and the operator link

Keep the IK, limits and stop on a computer on the robot's network, and let the operator's poses and the video cross the internet. Below roughly 100 ms round trip, hand-tracked grasping feels direct. Between 100 and 200 ms, slow the arm and accept fewer fine grasps per hour. Above that, walk and position the robot remotely and leave fine manipulation to a policy the operator supervises. Our latency guide has the budgets by task.

What a good Booster operator does

Moving from a T1 to a T2

Treat the T2 as a new robot type. Its body, arm kinematics, cameras and compute differ from the T1's, so T1 datasets and policies do not transfer directly. Locomotion policies trained in booster_gym for the T1 need retraining for the T2's body, and any teleop retargeting needs new kinematics. What carries over is the software pattern: the same SDK family, the same deploy-in-sim-then-real workflow, Booster Studio, and your operators' protocol. Keep T1 and T2 episodes in separate datasets with the robot type in the metadata. If you need a smaller, cheaper Booster for fleets, see the Booster K1.

Takeover and fallback

Once a T2 runs a policy, the custom teleop stack doubles as the takeover channel. Define what triggers a handover, keep the stop one action away, and record every takeover as a labeled correction episode for the next training run. Our human-in-the-loop guide covers the escalation design.

How RoboPilots runs Booster teleoperation

Your robot stays at your site. We agree the teleop stack with you, whether you have built one or need one assembled on the Booster SDK, set up the operator link through a computer on your network, and train a dedicated group of operators on your unit and your task protocol. Several humanoid makers have sent us robots to test our operating protocols, and that is the starting point for every new platform.