Dog Activity Trackers
Duke EGR 101Two Duke Engineering Foundry teams building wearable activity trackers for foster dogs at Saving Grace Animal Shelter. I served as Teaching Assistant, consulting on hardware design, firmware, and testing methodology.
Context
Dogs at Saving Grace Animal Shelter are frequently placed with foster families as part of rehabilitation programs for abused or at-risk animals. Full rehabilitation requires daily monitoring of specific activity and movement levels — but no affordable, subscription-free solution existed for the shelter’s budget.
Client Chip Bobbert engaged two separate Duke EGR 101 Foundry teams to tackle the problem. I served as Teaching Assistant, consulting both teams throughout the semester on hardware decisions, electronics design, firmware architecture, and testing methodology.
Team 1 — Dog Activity Tracking Device (Team MAAAC)
Alexandre Dias, Ari Dixit, Arshaan Sayed, Conrad Qu, Mila Prakapenka
This team built a GPS + accelerometer collar device with a dedicated ground station. The standout technical contribution was a Kalman filter fusing GPS position data and accelerometer measurements to produce a more accurate distance estimate than either sensor alone — assigning weights based on each sensor’s predetermined uncertainty characteristics.
Architecture
- On-collar device: GPS module + accelerometer, radio transceiver (HC-12)
- Ground station: HC-12 receiver + ESP32 with Wi-Fi, uploads to Firebase Realtime Database
- Frontend: HTML/CSS/JS web app — users set target activity levels and view the past two days of stats
Results
| Criterion | Target | Result |
|---|---|---|
| Cost | < $50, no subscription | Pass — $42.72 |
| Durability | Water & impact resistant | Pass — submerged 15 cm, dropped 3 m |
| Accuracy | < 10% distance error | Pass — +8.7% error |
| Weight | < 400 g | Pass — 123 g |
| Safety | No sharp edges | Pass |
| Battery life | ≥ 3 days | TBD |
| UI usability | Rating ≥ 3/4 from 10 users | TBD |
Team 2 — Dog Fitness Tracker
Sam Patterson, Pablo Garza T, Yiannis Lempidakis, John Button, Jacob Hills
This team took a simpler, lower-cost approach: an accelerometer-only collar clip storing distance travelled in a local memory log, transmitted wirelessly to a remote interface. The emphasis was on ease of use — the device was designed to be set up by a non-technical foster family with no prior experience.
Architecture
- On-collar device: IMU (accelerometer) + ESP8266, clips directly to collar
- Casing: 3D-printed enclosure with battery port seal for waterproofing
- Interface: Remote web interface for activity log review
Results
| Criterion | Target | Result |
|---|---|---|
| Cost | < $50 | Pass — $17.54 |
| Durability | Waterproof to 1 m, lasts > 1 year | Pass |
| Size | < 3 oz | Pass — 2.3 oz (65 g) |
| Safety | AKC regulations compliant | Pass |
| Data accuracy | < 15% error | Failed — 75% accuracy (25% error) |
| Battery life | > 2 days | Failed — ESP8266 drew more current than anticipated |
Consulting Role
Across both teams my involvement covered:
- Hardware selection — advising on MCU and sensor choices relative to power budget and cost constraints
- Firmware guidance — sensor fusion strategy (Kalman filter implementation for Team 1), data logging architecture for Team 2
- Testing methodology — helping teams design valid, repeatable test procedures for each design criterion
- Presentation coaching — working with Dr. Elizabeth Paley to prepare teams for final poster presentations
Both teams’ work contributed to Saving Grace Animal Shelter’s ongoing rehabilitation program for foster dogs.