← Back to Projects

Autonomous Agricultural Robot '26

Ground robot for precision farming and crop monitoring

Robotics ROS Autonomous Navigation Computer Vision
Precision Agricultural Robotics Club robot

Why

A ground-based autonomous platform built to perform precision farming tasks — crop health monitoring, weed detection and mapping, and targeted intervention — with minimal human input. Built on ROS, the robot fuses GPS, LiDAR, IMU, and camera data to navigate real crop rows reliably, even through partial GPS dropout or variable lighting.

The modular architecture means the same platform can swap tools depending on the task: survey pass, weed mapping, or soil sampling.

Which

🔧 Mechanical

  • Custom aluminum chassis · high ground clearance
  • Four-wheel independent drive
  • Weatherproof electronics housing
  • Modular sensor & tool mounting
  • Quick-swap 48V Li-ion battery

⚡ Hardware

  • NVIDIA Jetson Xavier NX
  • RTK-GPS · 360° LiDAR · stereo camera
  • RGB + multispectral cameras
  • IMU · brushless motors + encoders
  • 4G/LTE for remote monitoring

💻 Software

  • ROS Noetic · SLAM · sensor fusion
  • OpenCV crop row & weed detection
  • Path planning for full-field coverage
  • Web-based ground control station

Wrinkles

GPS Dropout in the Field

Challenge — Tree lines and hills degrade GPS, causing positioning failures mid-run.

Solution — Sensor fusion combines RTK-GPS, IMU, wheel odometry, and visual odometry. LiDAR-based localization serves as fallback during signal loss. Result: ±5cm open field, ±15cm during dropout.

Crop Row Detection Across Growth Stages

Challenge — Lighting variation, weeds, and crop growth stage all break naive vision approaches.

Solution — Adaptive thresholding + Hough transform for row edges; ML-based crop/weed classification. 94% detection accuracy from seedling through full maturity.

Battery Life for Full-Field Runs

Challenge — Motors, sensors, and onboard compute drain batteries fast — need 4–6 hour operational range.

Solution — Power-efficient motor control + dynamic sensor sampling + quick-swap battery design. Extended to 5+ hours; energy use down 30% vs initial design.

Wins

  • Navigation: ±5cm open field · ±15cm during GPS dropout
  • Crop Row Detection: 94% accuracy across types & growth stages
  • Obstacle Avoidance: 100% success in test scenarios
  • Coverage: 2–3 acres/hour at survey speed
  • Runtime: 5+ hours per charge
  • Scouting Labor: 70% reduction vs manual in field trials