Researchers tested whether a construction humanoid should follow human demonstrations or exploit its own embodiment once the task goal is clear. On an adult-sized Unitree H1-2, they compared human-derived motion, generated through vision-based capture and retargeting, with robot-specific motion, generated using inverse kinematics and minimum-jerk transfers to reach the same targets while departing from the demonstrated posture.

In a pick-and-stack sequence representing construction material handling, with four matched motion targets and three trials per condition, robot-specific motion was substantially more efficient: 12.87 s versus 25.03 s in execution time, 0.92 m versus 4.33 m in trajectory length, and 58.7 J versus 247.3 J in two-arm mechanical work. The departure from human posture was verified using Rapid Upper Limb Assessment (RULA).

The results suggest that allowing humanoids to move beyond human-derived motion could open new efficiency gains in construction tasks. However, the authors note that the study used a single task and a small number of trials, and that broader construction tasks are needed to determine whether such differences translate into real productivity improvements.