具身智能

03 / EXPLORE

Connect perception, decisions and actions to the physical world.

START HERE

Series overview · suggested reading order

  1. Tutorials

    Embodied AI Practice I: Pick up a cube in MuJoCo, then deliberately miss

    Run a contact-based pick-and-place task, compare0,25 and50mm target biases, and preserve videos, state trajectories and explicit success criteria.

    18 MINPublished
  2. Engineering cases

    Embodied practice II: if the grasp failed, skip the transfer

    An 18-rollout MuJoCo comparison of a lift/contact gate, with recorded trajectories, videos, runnable code and explicit limits.

    12 MINPublished
  3. Engineering cases

    Embodied practice III: a confirmed grasp still needs monitoring

    A 27-rollout MuJoCo comparison of transient observation loss and forced opening, with false cancellations, a 40 ms confirmation delay, code and recorded trajectories.

    14 MINPublished
  4. Engineering cases

    Embodied practice IV: messages arrive, but observations expire

    45 MuJoCo rollouts separate sampling intervals, bad-sample counts and capture age, showing how old messages can hide grip loss and why duration thresholds still have limits.

    14 MINPublished
  5. Engineering cases

    Embodied practice 5: communication is back—may transfer resume?

    18 MuJoCo rollouts reveal unsupported resumptions hidden by eventual success; validate fresh evidence and replan the remaining path.

    12 MINPublished
  6. Engineering cases

    Embodied AI VI: what do recovery gating and replanning each solve?

    A 30-rollout ablation reduces first target jumps from 50–97 mm to about 0.25 mm, without improving placement outcomes in this fixed matrix. Separate recovery evidence, command continuity and task success.

    14 MINPublished
  7. Engineering cases

    Embodied AI Practice VII: Why a Transfer Rule Stops Normal Lowering

    21 MuJoCo rollouts separate stage-specific grasp checks, stop actions and final placement. A false alarm and lingering finger contact reveal two different failure modes.

    13 MINPublished
  8. Engineering cases

    Embodied AI VIII: what should the gripper do after a drop alarm?

    Fifteen fixed-scene rollouts compare hold, open and open-then-retreat after identical alarms.

    9 MINPublished
  9. Engineering cases

    Embodied AI IX: a release command is not verified placement

    Apply single-capture and sustained-window checks to five actual trajectories, including forged observations and post-completion disturbance.

    9 MINPublished
  10. Engineering cases

    Embodied AI X: why can a recovery budget stop a recoverable task?

    Twelve rollouts compare unlimited recovery, a two-resume cap and a 400 ms cooldown, separating bounded behavior from task completion.

    9 MINPublished
  11. Engineering cases

    Embodied AI practice (11): placement completed, but is it still valid?

    Seven MuJoCo rollouts and a paired replay separate historical completion from current validity, including false alarms and expiry boundaries.

    11 MINPublished
  12. Engineering cases

    Embodied AI practice (12): what should the gripper do after recovery is exhausted?

    12 MuJoCo rollouts and synchronized three-view replay compare hold, open and open-retreat, separating terminal decisions from physical consequences.

    12 MINPublished
  13. Engineering cases

    Embodied AI practice (13): where should recovery resume after interrupted lowering?

    18 MuJoCo rollouts and synchronized replay connect phase contracts to recovery budgets, separating false holds, physical placement and controller abort.

    13 MINPublished
  14. Engineering cases

    Embodied AI practice (14): does landing in the bin mean the task is complete?

    Eight MuJoCo trajectories, three passive readouts and recorded-state replay separate historical placement, current evidence and qualified completion.

    13 MINPublished
  15. Mechanisms

    Reading OpenVLA — from an image and instruction to robot actions

    Follow preprocessing, visual projection, action tokens and unnormalization, with particular attention to units, boundary indices and the environment interface.

    12 MINPublished
  16. Engineering cases

    A second MuJoCo scene: why does the actuator overshoot?

    Compare mass and PD damping in a planar force-controlled actuator, separating target crossing from stable completion.

    9 MINPublished
  17. Engineering cases

    MuJoCo obstacle clearance: a point path is not a body path

    Six contact trajectories compare direct, edge and clearance paths while separating geometry from controller tracking.

    9 MINPublished
  18. Engineering cases

    Delayed MuJoCo feedback: can an old observation predict now?

    Twenty-four paired traces compare delay, position noise and constant-velocity prediction, retaining incomplete recoveries.

    9 MINPublished
  19. Engineering cases

    MuJoCo velocity estimation: smoother signals can mean worse control

    54 actual simulation rollouts compare finite differences, EMA and an oracle reference across position noise and observation delay.

    11 MINPublished
  20. Engineering cases

    MuJoCo controller tuning: best candidate is not a passing controller

    Select on 54 development episodes, then check position error and stable completion on 36 held-out episodes.

    10 MINPublished
  21. Engineering cases

    MuJoCo observation loss: zero force is not a stop

    Compare stale feedback, zero force and fading in 27 episodes to distinguish timeout from physical braking.

    10 MINPublished
  22. Engineering cases

    MuJoCo mass transfer: can the same gains still work?

    Compare fixed and mass-scaled gains in 36 episodes, including lower mean error without a velocity-contract pass.

    10 MINPublished

PROJECTS

Related projects

  1. PROJECT / 003

    Hohoo Embodied Experiments

    Two MuJoCo teaching scenes—pick-and-place and planar force control—record contracts, tuning, clearance and delayed-feedback outcomes, including failures.

    Experiment
Related experiments
  1. LAB / 004

    VL01 / Pickup target offsets

    What changes when pickup x targets shift by 25 / 50 mm with the controller fixed?

    Completed
  2. LAB / 006

    E2 / Confirm the grasp before transfer

    Use sustained height and bilateral contact to decide whether transfer should proceed.

    Completed
  3. LAB / 007

    E3 / Contact monitoring during transfer

    Separate missed loss, false cancellation and confirmation delay using forced opening and one empty contact report.

    Completed
  4. LAB / 008

    E4 / Observation freshness and sampling

    Separate delivery from fresh evidence and test a fixed count at different sampling periods.

    Completed
  5. LAB / 009

    E5 / Revalidating recovery

    Does delivery recovery authorize the old action? Compare fresh evidence, replay and remaining-path planning.

    Completed
  6. LAB / 012

    E6 / Recovery gate and path ablation

    Separate revalidation from replanning: which differences concern evidence and which concern the resumed target?

    Completed
  7. LAB / 013

    E7 / Grasp contracts during lowering

    What changes when transfer monitoring extends into lowering?

    Completed
  8. LAB / 016

    E8 / Post-alarm exit actions

    Fifteen fixed-scene rollouts compare hold, open and open-then-retreat after identical alarms.

    Completed
  9. LAB / 017

    E9 / Post-release completion verification

    Apply single-capture and sustained-window checks to five actual trajectories, including forged observations and post-completion disturbance.

    Completed
  10. LAB / 018

    E10 / Bounded recovery and cooldown

    Twelve rollouts compare unlimited recovery, a two-resume cap and a 400 ms cooldown, separating bounded behavior from task completion.

    Completed
  11. LAB / 020

    E11 / Completion and current validity

    Seven MuJoCo rollouts and a paired replay separate historical completion from current validity, including false alarms and expiry boundaries.

    Completed
  12. LAB / 023

    Embodied AI practice (12): what should the gripper do after recovery is exhausted?

    12 MuJoCo rollouts and synchronized three-view replay compare hold, open and open-retreat, separating terminal decisions from physical consequences.

    Completed
  13. LAB / 026

    Embodied AI practice (13): where should recovery resume after interrupted lowering?

    18 MuJoCo rollouts and synchronized replay connect phase contracts to recovery budgets, separating false holds, physical placement and controller abort.

    Completed
  14. LAB / 029

    Embodied AI practice (14): does landing in the bin mean the task is complete?

    Eight MuJoCo trajectories, three passive readouts and recorded-state replay separate historical placement, current evidence and qualified completion.

    Completed
  15. LAB / 036

    A second MuJoCo scene: why does the actuator overshoot?

    Compare mass and PD damping in a planar force-controlled actuator, separating target crossing from stable completion.

    Completed
  16. LAB / 037

    MuJoCo obstacle clearance: a point path is not a body path

    Six contact trajectories compare direct, edge and clearance paths while separating geometry from controller tracking.

    Completed
  17. LAB / 038

    Delayed MuJoCo feedback: can an old observation predict now?

    Twenty-four paired traces compare delay, position noise and constant-velocity prediction, retaining incomplete recoveries.

    Completed
  18. LAB / 041

    MuJoCo velocity estimation: smoother signals can mean worse control

    54 actual simulation rollouts compare finite differences, EMA and an oracle reference across position noise and observation delay.

    Completed
  19. LAB / 044

    Planar control tuning, observation loss and transfer

    153 episodes and 306000 physics steps. The selected controller passes 7/18 holdouts; zero force coasts, and mass scaling lowers mean error without passing the velocity contract.

    Completed

Paper reading guides

Related news

Published updates from external sources.

arXiv cs.RO · Original text · Translation unavailable

Robotic Data Inoculation ↗

Modern robot learning systems feed on high-quality demonstration data, and their performance hinges on coverage. Without good coverage, small errors compound and the policy can …

Original source

Browse all AI news →

Learning path & planned topics

These are future topics, not completed experiments.

  1. What is embodied AI? · PLANNED
  2. Understand perception, decisions and actions · PLANNED
  3. Read a paper with a question in mind · PLANNED
  4. VLM and VLA · PLANNED
  5. Action Space · PLANNED
  6. Embodied Benchmark · PLANNED

Continue along a learning path →