Repo
Isaac Lab → G1 locomotion
URDF, reward function, domain-randomization ranges, and the 48-hour sim-to-real workflow. Clone and run.
View on GitHub →Starting with the full Unitree range — Go2, B2, G1, R1, H1, H2 and arms. We make the platform do the job, end to end.
one short form · engineering review · reply within 1 business day
# unitree_sdk2_python — joint position control # Works across Go2, B2, G1, H1, H2, R1 from unitree_sdk2py.core.channel import ChannelFactoryInitialize from unitree_sdk2py.g1.arm.g1_arm_action import G1ArmAction ChannelFactoryInitialize(0, "eth0") # DDS over the robot's LAN arm = G1ArmAction() # Safety: zero-torque, then staged enable arm.setFsmLowCmd(); arm.send() arm.setFsmJoint(); arm.send() # compliant mode arm.startRayCasting() # perception hook
Platforms
4D LiDAR robot dog, ~15 kg, 2.5 m/s, climbs 30° — the accessible workhorse.
40 kg+ payload, IP-rated, all-terrain endurance.
23 DoF base / up to 43 DoF EDU, dexterous hands, indoor R&D platform.
≈ 123 cm, agile, developer-friendly entry humanoid.
1.78 m, fastest production humanoid (~3.7 m/s), high-torque.
Refines H1-2 — unified compute, stronger arms, longer battery.
Dual-arm teleoperation rig, <100 ms latency, builds training datasets.
6+ DoF arms for cells and benchtop automation.
Services
01 — Sell
We supply the right Unitree platform, configured for your application, with the spares and accessories to actually use it.
02 — Program
We write the code that makes the platform do the job — unitree_sdk2, ROS 2, Isaac Lab, LeRobot. Same SDK the Unitree teams use.
03 — Integrate
End-to-end integration — hardware, mounting, networking, perception stack, on-site commissioning, and operator training.
04 — Security & Hardening
Post-UniPwn (2025) — verified firmware, Bluetooth lockdown, isolated networks, and continuous monitoring. Legitimate hardening, never circumvention.
How it works
Step 0 — run it before step 01 to know what you actually need.
Open the tool →
Five-question form, three minutes. We learn what you're trying to do.
We name the platform, the SDK, the integration pattern, and a realistic timeline. Free, no commitment.
Fixed price. Plain-English SOW. You own the program source, the design, and the documentation.
We program in sim, validate on hardware, deploy on your floor. Remote support after, if you want it.
Stack
SDK
unitree_sdk2 / unitree_sdk2_python
C++ and Python SDK across the full Unitree range.
Middleware
ROS 2 (unitree_ros2)
Same DDS interface on real hardware and in simulation.
Simulation
Isaac Lab (unitree_sim_isaaclab)
High-fidelity sim-to-real workflow, validated policies.
Learning
LeRobot + UnifoLM-VLA-0
Imitation learning and Unitree's open VLA model.
Perception
LiDAR SLAM · RealSense · CUDA pipelines
Localization, mapping, multi-sensor fusion.
Safety
ISO 10218 · RIA R15.06
Industrial robot safety standards we scope to.
Security
Firmware hardening · network isolation
Post-UniPwn 2025 — verified patches, isolated robot networks.
Hardware
EOAT · power · mounting
Grippers, batteries, charging, custom mounts.
Why qtvue
qtvue exists because mid-market operators and research teams don't need another hardware vendor — they need someone who can make the platform do the job. We work inside the same public SDKs the platform teams publish, in the same simulators, on the same standards. When you hire us, you get a programming and integration team that ships.
We are pre-launch. We don't have a 240-cell case-study library to point at — and we won't fake one. What we do have is the depth to tell you, in plain English, what your use case will actually take. Some of them we will talk you out of. Most we will scope honestly.
Honest spec callouts
The kinds of specifics that signal we actually know the hardware.
47 checks · 3 layers · saves locally · no login.
Open the tool →
Proof
We're pre-launch, so we won't point you at a fake case-study library. Instead, here's runnable source and real measurements behind the work — everything we publish about is verifiable.
Repo
URDF, reward function, domain-randomization ranges, and the 48-hour sim-to-real workflow. Clone and run.
View on GitHub →Repo
50 teleoperated demonstrations to a deployed Go2 policy. Dataset format, config, eval script.
View on GitHub →Measurement
We instrumented the G1, Go2, and B2 under realistic load. The marketing numbers are standby figures.
Read the data →Five questions, three minutes. We name the platform, the SDK, the integration pattern, and a realistic timeline. No sales call required.
or just need programming help on a platform you already own? Same form.