Active stereo-camera study benchmarks sensor mixes on Unitree G1
Evidence notes
- The paper benchmarks 14 sensor combinations for ACT imitation learning on a Unitree G1 humanoid with three-finger hands.
- It reports that a minimal active stereo-camera setup outperformed more complex multi-sensor configurations in data-limited manipulation tasks.
- The benchmark gives G1 a concrete manipulation study on sensor choices for imitation learning.
Company context
Develops quadruped and humanoid robots, with strength in dynamic locomotion, vertically integrated hardware, and relatively low-cost commercial deployment. Unitree is one of the clearest examples of a legged robotics company moving from research visibility into real productisation and broader market distribution.