Low-cost servos make a small walking robot affordable, but they also make simulation unusually unforgiving: a policy can look stable in MuJoCo and fail when real actuator behavior differs. The project’s most useful contribution is therefore not just a printable robot design, but a practical path for connecting a modeled, learned gait to hardware.
Embodied AI2024
Open Duck Mini
Builds a roughly 42 cm bipedal robot designed to cost under $400 in parts. Its locomotion policies are trained in MuJoCo and tested on the physical robot, making sim-to-real transfer the central challenge.
Introduction
More Items
Rewinds world-model trajectories to test alternative robot actions and turn outcome comparisons into action-level training signals. Real-robot data continually corrects the model; across three physical tasks, success rates beat the strongest baseline by 10.67–11.85 percentage points.
Conditions robot action generation on predicted visual and motion features, capturing physical dynamics without generating future video during control. Reports results across four simulation benchmarks.
AI Dataset2026
Provides continuous human whole-body manipulation demonstrations for robot policy training, including locomotion and coordinated two-hand actions. Motion is reconstructed from egocentric observations, with reported whole-body joint error of about 3 cm.