agri rover

Autonomous Agri-Rover uses edge AI computer vision and real-time soil telemetry for automated crop health monitoring and targeted smart irrigation, reducing water waste by 30%.

Comment

Inspiration

Smallholder farming is severely constrained by unmonitored resource waste, delayed crop disease identification, and labor-intensive manual field checks. We were inspired to build the Next-Gen AI Agricultural Rover to bring low-cost, real-time edge intelligence directly to small-scale agriculture, automating crop diagnostic feedback loops without requiring expensive industrial machinery or constant cloud connectivity.

What it does

The Agri-Rover is an autonomous, dual-module robotic system that continuously observes, diagnoses, and acts in real-time:

  • Sense & Telemetry: Capacitive soil sensors gather live moisture levels while environmental safety sensors monitor for gas, flame, and motion risks.
  • See (Edge AI Vision): An onboard camera module powered by the XIAO ESP32S3 Sense executes machine learning vision inferencing locally to detect crop health and disease indicators without cloud latency.
  • Act (Targeted Irrigation): When moisture thresholds drop or crop stress is flagged, an integrated pump relay triggers micro-targeted irrigation directly where the plant needs it, saving up to 30% water.

How we built it

  • Architecture: Designed a dual-controller layout dividing tasks between high-level vision inferencing and low-level actuation.
  • Hardware Stack: Powered by a Seeed Studio XIAO ESP32S3 Sense module (vision/AI) and an ESP8266 controller paired with high-torque MG995 metal-gear servos, motor drivers, and an SSD1306 OLED readout.
  • Power Management: Engineered a stable power rail using a 2S 18650 Li-ion battery pack protected by a 2S BMS and regulated through a step-down buck converter to prevent brownouts under high servo loads.

Challenges we faced

  1. Power & BMS Tripping: High initial current spikes from MG995 metal-gear servos caused voltage drops, tripping the BMS under load. We resolved this by integrating step-down buck voltage regulation and dedicated rail filtering.
  2. Logic Level Signal Limits: Solved L298N driver compatibility and PWM syntax constraints to ensure smooth motor and servo control from 3.3V logic outputs.
  3. Edge Optimization: Running vision processing locally on low-power micro-controllers required optimizing model size for smooth on-device execution.

What we learned

  • Deepened hands-on expertise in multi-rail power distribution, battery balancing, and protection circuitry for robotics.
  • Optimized micro-edge computer vision pipeline integration using ESP-NOW / low-latency microcontroller protocols.
  • Mastered modular mechanical and electrical separation for robust field deployment.

What's next for Agri-Rover

  • Integrating autonomous pathfinding, GPS waypoint navigation, and long-range LoRa wireless telemetry.
  • Conducting pilot test deployments across local automated greenhouses and smallholder farming cooperatives.

Built With

Share this project:

Updates

Submission history