Q-Aid is a clip-on wearable AI companion badge built entirely around the Arduino UNO Q's dual-brain architecture — the Qualcomm QRB2210 running local AI on Linux, paired with the STM32U585 MCU for real-time sensor control. Everything runs fully on-device, with zero cloud dependency.
THE IDEA
Most "smart" wearables today rely on the cloud to think. Q-Aid flips that: every mode — voice processing, obstacle detection, translation, fall detection — runs locally on the badge itself. That means it keeps working with no Wi-Fi, minimal latency, and no data ever leaving the device unless the user chooses to sync it.
FOUR SWITCHABLE MODES
- Memory — a voice diary with spoken reminders and photo capture, with optional Wi-Fi sync
- Assist — ultrasonic obstacle detection to help visually impaired users navigate safely
- Translate — real-time, on-device speech translation
- Companion — environmental monitoring, step counting, fall detection, and a local SOS alarm, shown on an expressive OLED "face"
The user switches between modes with a single tap on a capacitive touch zone built into the enclosure.
HARDWARE
- Arduino UNO Q (main board)
- INMP441 I2S MEMS microphone
- MPU6050 (motion) + AHT20 (environment) sensors
- HC-SR04 ultrasonic distance sensor
- SSD1306 OLED display
- 8MP mini camera module
- TTP223B capacitive touch sensor
- TP4056 charging module + LiPo battery
DESIGN
The enclosure is fully modeled in Autodesk Fusion, with circular cutouts for the camera and ultrasonic sensor, a clear OLED window, ventilation grilles, and a flush capacitive touch zone so the badge stays clip-on-friendly and pocket-sized.
CHALLENGES SO FAR
- Scope creep — my first feature list tried to do too much, including a GPS "find my badge" function before I realized the UNO Q has no cellular modem or GPS onboard. I split anything that needs a paired phone into a future roadmap instead of overpromising a device that isn't actually standalone.
- Form factor — I originally designed this around a desk companion robot before realizing a "worn, clip, or carry" wearable badge made far more sense for daily use. Redesigning the enclosure from a robot shape into a flat clip-on badge meant starting the Fusion model over, but the result is a much stronger, more wearable design.
- Touch vs. button — I initially planned a physical push-button for mode switching, then switched to a flush capacitive touch pad instead. Small change on paper, but it meant redesigning that whole section of the shell to remove the drilled hole entirely — a sealed capacitive pad doesn't need one.
- Component cost — some of the sensors I first specced were proprietary modules from a single supplier and cost more than they needed to. Swapping to generic, widely available I2C parts (MPU6050, AHT20, SSD1306, etc.) cut the electronics budget by more than half with zero loss of functionality.
- PCB assembly — I don't have soldering experience yet, so rather than assume I'd hand-solder the custom shield, I'm planning to prototype everything on a breadboard first to validate the wiring and firmware, then use PCBWay's assembly service for the final board.
WHERE THINGS STAND
The Fusion enclosure design is complete. Component selection is finalized, and the build is moving into assembly and firmware integration for each of the four modes. Remaining challenges: finalizing the shield's schematic/PCB layout and writing the on-device AI pipelines (vision, speech, translation) for the QRB2210 side of the board. More build photos and a working demo video coming soon!
