Overview
Conventional agricultural UAV workflows often require a single drone to complete scanning and spraying sequentially or rely on offline image processing, increasing mission time and reducing operational efficiency. This project addresses these limitations by introducing a decentralized dual-drone system consisting of:
- Scan UAV: Performs autonomous field scanning and onboard crop anomaly detection using an NVIDIA Jetson Orin and an HSV-based computer vision pipeline.
- Spray UAV: Receives target GPS coordinates in real time, autonomously navigates to the detected locations, and performs precision spraying using vision-assisted target centering.
Drones Design
Scan Drone
The scanning drone frame is built using two 3D-printed ABS plates connected by 16 mm square carbon-fibre rods. A stainless steel plate is added at high-load regions to handle structural stresses during flight. An F550 landing gear is used to support safe takeoff and landing. The frame includes custom 3D-printed mounts for mounting the camera and flight computer, ensuring proper alignment and mechanical support.
Spray Drone
The spraying drone frame is constructed using carbon-fiber rods positioned between two carbon-fiber plates. The propulsion system consists of X6 motors powered by two 6S batteries connected in series. The landing gear is built using aluminium rods with custom mounts, providing ground clearance and structural support. A 10-litre pesticide tank is mounted between the landing gear legs.
System Architecture
Spraying Mechanism
The spraying system was designed to support both precision and wide-area applications using a dual-nozzle configuration. Solid-cone axial-flow nozzles were selected because they provide uniform droplet distribution and effective canopy penetration. A 45° cone-angle nozzle was used for precision spraying, while a 120° cone-angle nozzle provided wider area coverage. The nozzles were mounted on opposite ends of the landing gear and oriented inward to ensure accurate spraying at the center of the target plant. It is also equipped with an adaptive dual-nozzle spraying system, enabling dynamic selection between different spray patterns during operation. The spraying maneuver is controlled using a custom PCB controlled using a Raspberry Pi 5, which actuates relays for nozzle selection and pump control. The system integrates both manual triggering via an RC switch and autonomous operation based on geotagged target location received from the scan drone. On reaching the target location, the onboard controller deploys the selected nozzle followed by the pump to initiate spraying.
Electronics Architecture
The dual-UAV system consists of two independently operating drones connected through real-time telemetry. The Scan Drone uses an NVIDIA Jetson Orin Nano for onboard crop detection, while the Spray Drone uses a Raspberry Pi 5 to control the spraying mechanism. Both drones are equipped with a CubeOrange+ flight controller, Here 4 GPS, SIYI A8 Mini gimbal, and telemetry modules for autonomous navigation, communication, and coordinated precision spraying.
Communication Architecture
The dual-UAV system uses a 433 MHz telemetry link for direct communication between the Scan Drone and Spray Drone, enabling real-time transmission of target coordinates. Both drones are independently connected to the Ground Station via 915 MHz telemetry, allowing simultaneous mission monitoring, status updates, and manual intervention when required.
Results
The proposed dual-UAV system was successfully validated through real-world experiments on a 2-acre agricultural field. The onboard vision pipeline achieved ~95% detection recall with minimal false positives under varying lighting conditions. Reliable decentralized communication was maintained between the Scan Drone and Spray Drone over distances of up to 300 m, providing an initial target localization accuracy of 2–3 m. Using the vision-assisted target centering algorithm, the Spray Drone further refined its position to a steady-state error of 5–20 cm, enabling highly accurate and efficient precision chemical application.
Members
- Spruha Kshirsagar
- Rajasi Deshmukh
- Harshal Kolhe
- Samiron Biswas
- Mrunmayee Limaye
- Kumar Ayush
- Sanchet Dhalwar
- Ishani Khaty
- Abhinav Anand
- Sunil Watgule
- Vishnu Swaminathan
- Tarun Kayala