Robot Data and Hardware

Nav2 RateController keeps ticking a running child between nominal replanning intervals. Distinguish tree ticks, action requests and completed paths before tuning hz.

Choose Nav2 velocity-smoother feedback by comparing the last commanded speed with fresh odometry at the smoothing cycle. Check tracking, delay and stale-command behavior.

AnySkin makes a tactile surface replaceable, but policy reuse still needs a matched check. Separate unloaded readings, contact response and task transfer.

VRB extracts contact and hand-motion cues from human video. Learn what that supervision contributes and what the robot interface must still supply.

Touch100k preserves short tactile-property phrases and richer contact descriptions. Match each label level to the query and evaluate them on separate, object-aware splits.

UMI-style handheld demonstrations need more than video. Check relative gripper trajectories, latency alignment, two-hand pose and robot feasibility before policy training.

A tactile sensor helps only when it detects the contact cue your robot can still act on. Use TacO’s task comparisons to design a local sensor trial.

A human-video pipeline can fail in reconstruction, robot grounding or policy learning. Compare clean, corrupted and model outputs under fixed downstream conditions to find the first bad handoff.

A kinematically matched leader may help repeated robot demonstrations, but its value depends on the task and the calibration still required. Use a pilot to count usable episodes.

Separate DROID camera-to-base and camera-to-camera corrections, read their quality filters, and decide which derived 3D artifacts need rebuilding.