20. Project Presentations

JeLamp — AI-Powered Gesture-Control Robotic Arm Lamp

This is the final week of Fab Academy 2026 — final project due June 4, 2026. I present JeLamp — an AI-powered gesture-control robotic arm lamp that combines a 3-DOF mechanical arm, onboard camera vision, custom electronics, and a Wi-Fi control interface.

JeLamp — AI-powered gesture-control robotic arm lamp, final project for Fab Academy 2026 at Chaihuo Makerspace, Shenzhen.

Summary Slide

presentation.png (1920 × 1080) — click to open full size. © 2026 Ning Zhang — CC BY-NC 4.0.

Presentation Video

presentation.mp4 — ~1 minute overview of JeLamp: AI gesture control, arm build, and demo.

Final Project Summary

What does it do?

JeLamp is an AI-powered gesture-control robotic arm lamp. An onboard camera runs a machine-learning hand-gesture model (rock–paper–scissors via SenseCraft AI) and maps detected gestures to coordinated 3-DOF arm motion and NeoPixel feedback. Users can also drive joints and lighting manually through a Wi-Fi web interface.

What did I design?

• 3-DOF mechanical arm structure (Fusion 360, 3D printed + laser cut)
• JeLamp carrier PCB — power, servo PWM headers, NeoPixel output (KiCad)
• ESP32-C3 NeoPixel driver PCB — I2C slave LED controller (KiCad)
• Hello World test board — XIAO ESP32-C3 + button + LED (KiCad, milled)
• Embedded firmware — gesture ML, I2C protocol, servo PWM, web server

What processes were used?

What worked? What didn't?

How was it evaluated?

Success criteria from Week 18: custom PCB with input + output functional, gesture recognition > 80% accuracy, NeoPixel responds via I2C, mechanical motion through ≥ 3 poses, system independently operable from single power supply, complete documentation with BOM and presentation materials.

System Integration

JeLamp integrates a 3-DOF robotic arm, custom carrier PCB, embedded AI vision, and a web interface into one lamp product. The architecture evolved from the dual-MCU plan documented in Week 16 toward a single JeLamp carrier PCB that hosts the XIAO ESP32-S3 Sense, power regulation, servo PWM headers, and NeoPixel output — reducing wiring complexity while keeping subsystems modular for testing.

                    +---------------------------+
                    |   User (hand gesture)     |
                    +-------------+-------------+
                                  |
                                  v
+------------------------------------------------------------------+
|  XIAO ESP32-S3 Sense  (on JeLamp carrier PCB)                    |
|  - OV2640 camera + SenseCraft AI gesture model                   |
|  - PWM servo output (3× joints)                                |
|  - NeoPixel output (direct or via ESP32-C3 I2C slave)            |
|  - Wi-Fi web UI (Week 15)                                        |
+----------------------------+-------------------------------------+
                             |
         +-------------------+-------------------+
         v                   v                   v
   PWM servos          NeoPixel ring       HTTP web client
   (base/elbow/head)   (gesture feedback)  (manual control)

Integration flow:
  Camera -> SenseCraft AI inference -> gesture label
       -> 3-DOF arm pose preset + NeoPixel status color
       -> (optional) I2C to ESP32-C3 lighting board

Individual subsystems were validated in weekly assignments before integration: ultrasonic input (Week 9), NeoPixel output (Week 10), wireless control prototype (Week 11), PWM servo control (Week 15), and gesture ML (Week 17). The remaining integration challenge is mechanical — routing power and signal cables through rotating joints without binding.

Bill of Materials (BOM)

Full materials list for the JeLamp prototype. Detailed sourcing notes are on the Final Project page and in Week 18.

Costs are approximate prototype quantities; actual receipts tracked in Week 18 documentation.

Original Design Files

All original design files are archived on this website under public/files/ — no external-only hosting. Source code and board files are linked from the relevant weekly assignment pages.

KiCad source for the JeLamp carrier PCB and Fusion 360 mechanical files are documented step-by-step on the Final Project page. Full project archive is mirrored on GitLab.

License

Acknowledgements

• LeLamp (github.com/lelamp) — open-source robotic arm lamp; joint layout and mechanism reference.
• SenseCraft AI / Seeed Studio — pretrained hand-gesture model and XIAO ESP32-S3 Sense hardware platform (core AI input for gesture control).
• Edge Impulse — facial expression ML workflow explored in Week 17.
• Fab Academy — Neil Gershenfeld and the global instructor network; curriculum structure and documentation standards.
• Chaihuo Makerspace, Shenzhen — local Fab Lab: KEXU CNC, laser cutter, 3D printers, soldering station, and instructor support.
• Group assignment collaborators — Week 12 vending machine team for shared mechanical design experience.
• KiCad, Fusion 360, Arduino ESP32 core — open-source and industry-standard design tools used throughout the project.

Related Documentation

Presentation Schedule

Final presentations for Fab Academy 2026: finalprojects.fabacademy.org — 2026 schedule