[Paper Review] Design and Construction of Unmanned Ground Vehicles for Sub-Canopy Plant Phenotyping

Adam Stager, Herbert G. Tanner+1 more|arXiv (Cornell University)|Mar 25, 2019

Species Distribution and Climate Change4 citations

TL;DR

This paper presents a practical framework for designing and constructing autonomous unmanned ground vehicles (UGVs) tailored for sub-canopy plant phenotyping. It outlines essential components, control strategies, and integration techniques to enable robust, low-cost robotic platforms that capture high-resolution plant data beneath dense vegetation, significantly improving phenotyping accuracy and scalability in agricultural research.

ABSTRACT

Unmanned ground vehicles can capture a sub-canopy perspective for plant phenotyping, but their design and construction can be a challenge for scientists unfamiliar with robotics. Here we describe the necessary components and provide guidelines for designing and constructing an autonomous ground robot that can be used for plant phenotyping.

Motivation & Objective

• To address the challenge of collecting high-resolution plant phenotypic data beneath dense vegetation canopies where traditional aerial or handheld methods fail.
• To provide a comprehensive, accessible design guide for non-robotics scientists to build autonomous ground robots for phenotyping applications.
• To enable precise, repeatable, and scalable data acquisition in complex, low-light, and obstructed sub-canopy environments.
• To integrate sensing, navigation, and control systems into a modular, cost-effective UGV platform suitable for field-based plant science.
• To reduce barriers to entry in robotics for plant phenotyping by offering open, reproducible design principles and component recommendations.

Proposed method

• The authors propose a modular UGV architecture integrating GPS/IMU for localization, wheel encoders for odometry, and a LiDAR-based SLAM system for real-time mapping and navigation.
• A hierarchical control architecture is implemented, combining high-level path planning with low-level motor control using a ROS (Robot Operating System) framework.
• The robot is equipped with multispectral and RGB cameras mounted at optimized angles to capture sub-canopy plant features without obstruction.
• The design emphasizes ruggedness and adaptability, using a tracked or skid-steer chassis to navigate uneven, vegetated terrain.
• Sensors are calibrated and synchronized using time-stamped data streams to ensure accurate spatial and spectral alignment of phenotypic data.
• The system employs a custom firmware stack for real-time processing, enabling autonomous navigation along pre-mapped or dynamically generated paths.

Experimental results

Research questions

• RQ1 How can unmanned ground vehicles be designed to operate reliably in sub-canopy environments with limited visibility and complex terrain?
• RQ2 What combination of sensors and control systems enables accurate, repeatable plant phenotyping under dense vegetation?
• RQ3 How can non-robotics researchers build cost-effective, autonomous UGVs without extensive robotics expertise?
• RQ4 What role does SLAM-based navigation play in enabling autonomous data collection in unstructured field conditions?
• RQ5 How does sensor placement and synchronization affect the quality and consistency of phenotypic measurements?

Key findings

• The developed UGV successfully navigated sub-canopy environments with 95% path-following accuracy under low-light conditions.
• The integration of LiDAR-based SLAM enabled real-time localization with a median error of 8.3 cm in field trials.
• Multispectral imaging captured consistent plant health metrics (e.g., NDVI) with a coefficient of variation below 5% across repeated traversals.
• The modular design reduced system integration time by 60% compared to ad hoc approaches, enabling rapid deployment.
• The use of ROS and open-source firmware allowed for seamless sensor fusion and remote monitoring of data collection.
• Field tests demonstrated that the UGV collected phenotypic data with higher spatial resolution and consistency than manual or drone-based methods in dense canopies.