Smart Pet Potty Training System
The problem
First-time dog owners get reinforcement timing wrong. Operant-conditioning research is unambiguous: a reward delivered more than ~2 seconds after the desired behavior is interpreted by the animal as a reward for whatever they're doing now. Most owners can't be standing there, treat in hand, every time their puppy uses the right spot — and the misses pile up into a slower, more frustrating training cycle.
The system
Weight + moisture sensors under the designated potty pad → Arduino Nano arbitrates (suppressing false positives from play) → ESP32-coupled treat dispenser releases a single reward in under 2 seconds → owner sees a live confirmation on their phone via Blynk. The hardware sits flush in the corner of a room and disappears from daily life.
The HCI lens
The primary user is not the dog — it's the owner who is failing at consistent reinforcement. Every design choice was made for the human-side experience: zero-touch operation, ambient form factor, a phone notification that confirms "the system did the right thing for you", and treat refills measured in weeks rather than days. The dog's rewards are the system's output; the owner's confidence is the system's job.
Design process
Research
6 first-time dog-owner interviews. Pattern: not knowledge gap, but missed reward windows. Designed for presence, not education.
Ideate
3 concepts. Phone-buzz + camera both lost on the 2-second latency constraint. Autonomous dispenser concept won.
Prototype
Breadboard rig validated weight detection alone. Moisture sensor added for the dog standing on pad ≠ dog using it false-positive case.
Evaluate
12-week-old puppy, 3 weeks. End state: ~90% in-window rewards, 0 false rewards. Surprise win: owners trusted it enough to leave the house.
By the numbers
What I'd do differently next time
The biggest miss was treating the dispenser as a hardware problem rather than a UX one. The first revision jammed on a softer treat brand owners actually buy. A v2 would start the design from "what treat shapes does the typical owner already have in the cabinet" — a true human-centered constraint — and the mechanism would follow from that.