Automated Cat Litter Box

Overview

A five-person mechatronics project: a tumbling-drum litter box that separates clean litter from waste automatically. My work covered the electrical system, ESP32 firmware, and sensor/actuator integration — the mechanical design, CAD, and structural analysis were led by a teammate.

Team Project & My Contribution

Automated Cat Litter Box was a five-person mechatronics project. Mechanical design — the drum, drivetrain, and enclosure CAD and structural analysis — was led by a teammate. My work focused on the electrical system, firmware, and sensor and safety-logic integration.

• Camera-based detection: integrated and programmed the OpenMV vision system (motion-based ROI detection) and its serial link to the ESP32.
• Embedded control: wrote the ESP32 firmware and state machine (idle → cat-present → exit-delay → cleaning → fine-homing).
• Electrical integration: contributed to system wiring and sensor integration — load cell, Hall-effect sensor, current sensor — alongside the team.
• Safety logic: implemented the dual-sensor (load-cell + vision) presence check that immediately halts the cleaning cycle if the cat re-enters mid-cycle.

System Architecture

Testing & Results

• Load-cell detection: a clear separation between empty (~14,000–16,000 raw) and occupied (~45,000–47,000 raw) readings, giving a wide margin against false triggers.
• Vision detection: reliable under normal indoor lighting, but accuracy dropped under reduced or low lighting — this is why the load cell was treated as the primary detection layer, with the camera as a secondary check rather than the sole trigger.
• Hall-effect homing: consistent magnet detection across repeated cycles for repeatable drum positioning.
• Safety interlock: verified that re-entry during an active cleaning cycle immediately stops the motor and returns the system to idle.
• Structural validation (mechanical team): simulated stresses in the drum and base stayed well below the material's yield strength under expected operating loads.

Challenges & Limitations

• Vision performance was lighting-dependent; sensor fusion with the load cell — rather than vision alone — turned out to be the more robust detection strategy.
• The prototype was built at 50% scale via FDM 3D printing, so tolerances tuned for injection molding needed rework (wall thickness, mating clearances) to fit the printed parts.
• Long-term durability of the 3D-printed components is limited compared to the production-grade ABS intended for a manufactured version.