13 réponses à ce sujet

[ethanchenplus]

Hi everyone! I'm Kewei Chen.

My technical background: I am interested in robotics, embodied AI, robot manipulation, and edge AI deployment. I have experience working with real robotic arms, AI-assisted robotics workflows, and project documentation on GitHub.

My robot/AI dream: I plan to design an RDK X5-powered smart household inventory assistant with a real robotic arm. The system will use RDK X5 for on-device AI perception, recognize household supplies, manage inventory records, generate low-stock reminders, and trigger robotic arm actions for simple item interaction, sorting, or demonstration tasks.

Excited to join the challenge and learn with everyone! #RoboticsDreamKeeper #RDKX5

[ethanchenplus]

Stage 1 Update - 2026-05-21

Progress:
- RDK X5 / Magic Box board bring-up completed.
- SSH access, Wi-Fi networking, XFCE desktop capture, microphone recording, MIPI camera, and BPU YOLO inference verified.
- Public GitHub repository created.
- Official showcase PR submitted.

Evidence:
- Repository: https://github.com/S...inventory-robot
- Stage 1 package: https://github.com/S...1_SUBMISSION.md
- Official PR: https://github.com/D...hallenge/pull/1

Next:
- Continue toward Stage 2 system design and robotic arm integration.

[drobo_community]

Wow, you are already on the way!

[rustedfriend]

That's rad, i've considered building something similar for my shop and still want to eventually. I look forward to seeing your project.

[drobo_community]

Hi @kewei how's everything going?

[drobo_community]

Congratulations, you have unlocked your RDK Explorer title!

[ethanchenplus]

Stage 3 Final Update - 2026-07-15

TuntunClaw RDK X5 is now complete!

The final project combines our completed MuJoCo inventory-manipulation workflow with a real RDK X5 Magic Box and robotic-arm experiment.

Final results:
- OpenClaw natural-language task planning
- VLM + SAM object understanding
- GraspNet grasp-pose inference
- Persistent inventory and location memory
- RDK X5 BPU perception at 30.02 FPS
- Real robotic-arm retrieval with tablet inventory feedback
- Magic Box low-stock voice notification

In the physical experiment, the robot retrieves one Oreo and one Nestle coffee stick. Oreo inventory changes from 5 to 4 and remains above its threshold of 2, so no warning is issued. Coffee inventory changes from 7 to 6, reaching its threshold of 6, which triggers the replenishment warning.

Final demo:


GitHub:
https://github.com/E...inventory-robot

Official Showcase PR:
https://github.com/D...hallenge/pull/9

#RoboticsDreamKeeper #RDKX5 #Robotics #EmbodiedAI

[ethanchenplus]

@:robot:Lisa Thank you for following up and for the RDK Explorer title! The project has now completed its Stage 3 real-world experiment and final submission. I really appreciate your encouragement and the support from the D-Robotics community.

@RustedFriend (Isaac) Thank you! A shop inventory assistant is also a great use case for this workflow. The combination of inventory memory, low-stock alerts, and robotic manipulation could transfer naturally to a small shop or workshop. I'd be happy to exchange ideas and implementation experience with you.

[ethanchenplus]

Here is a separate look at our performance on physical robots

[ethanchenplus]

Repository migration update:

The original GitHub account became inaccessible to anonymous reviewers, so the public project repository and official Showcase PR have been migrated.

New public repository:
https://github.com/E...inventory-robot

New official Showcase PR:
https://github.com/D...hallenge/pull/9

The project content, commit history, Stage 1/2/3 evidence, and final demo remain unchanged.

[ethanchenplus]

...

[shockley6668]

HI @kewei, do you have an English version of this video? Since all submitted materials for this competition must be in English, an English version would be highly appreciated. Thank you!

[shockley6668]

Hi, @kewei I've reviewed your PR and requested changes on GitHub. Please add the reproducible robotic-arm and low-stock project implementation, complete the Quick Start, dependencies, BOM and Stage 2 engineering documentation, and replace the Chinese demo with an English version before requesting another review.

[ethanchenplus]

Hi @Shockley , thank you for the detailed review.

I have now addressed all requested changes, including the reproducible robotic-arm and low-stock pipeline, complete Quick Start and dependency documentation, the final BOM and Stage 2 engineering package, RDK fixed-camera visual delivery confirmation, and the English-narrated final demo.

The PR has been updated and I have requested another review on GitHub.

English final demo:


Showcase PR:
https://github.com/D...hallenge/pull/9
If you have any further questions or need any additional information, please feel free to let me know.
Thank you again!