We are proud to announce an advancement in agricultural robotics (specifically in horticulture) with the publication of research in IEEE Robotics and Automation Letters. Our latest work, “Navigating Narrow Spaces: A Comprehensive Framework for Agricultural Robots,” addresses one of the pressing challenges facing autonomous agricultural wheeled robots today.

The Challenge of Constrained Environments

Agriculture presents unique challenges for robotic systems. Unlike controlled indoor environments, agricultural robots must navigate through narrow spaces between crop rows, adapt to changing plant growth, and operate reliably in dynamic outdoor conditions. The confined spaces of polytunnels and glasshouses, where many high-value crops like strawberries are grown, demand exceptional precision and adaptive control.

Traditional navigation approaches often struggle in these environments, where a deviation of just centimetres can result in crop damage or system failure. Our research team, working within the Agri-OpenCore project, recognised that a new approach was needed.

A Modular Solution for Real-World Deployment

Led by researchers Geesara Kulathunga, Abdurrahman Yilmaz, Zhuoling Huang, Ibrahim Hroob, Jaspreet Singh, Leonardo Guevara, Grzegorz Cielniak, and Marc Hanheide, the team developed a modular and perception-driven navigation framework specifically designed for constrained agricultural environments.

The innovative approach integrates several key components:

• Multi-step point cloud processing: A robust pipeline for local perception that can adapt to the complex and variable nature of agricultural environments
• Intelligent structure detection: Advanced algorithms that identify structural elements within polytunnels and glasshouses
• Dynamic boundary estimation: Real-time analysis of navigable space boundaries
• Adaptive trajectory refinement: Continuous adjustment of robot paths based on detected environmental constraints

Proven Performance in the Field

The framework was rigorously tested in a real strawberry polytunnel, demonstrating superior performance compared to existing state-of-the-art navigation systems. In summary:

• Average lateral deviation of just 0.08±0.01 metres – achieving remarkable precision in narrow growing environments
• Consistent performance under increased velocity constraints – maintaining accuracy even during faster operations
• Superior trajectory accuracy and control stability compared to established methods including the Resilient Timed Elastic Band (RTEB) and Model Predictive Path Integral (MPPI) approaches

Advancing the Agri-OpenCore Vision

This research forms a crucial component of the Agri-OpenCore project, an ambitious initiative funded by Innovate UK to create the world’s first open development platform for agricultural robotics. Agri-OpenCore aims to revolutionise the agricultural sector by providing standardised access to core software and hardware components, enabling rapid adoption of robotic solutions across the industry.

The navigation framework developed by our team directly addresses Agri-OpenCore’s objective to “cut the time and cost to develop a robotic harvesting system for any farm/crop with human-cost-picking-parity performance.” By providing reliable navigation capabilities for narrow spaces, we’re removing one of the fundamental barriers to widespread agricultural robot deployment.

The full paper, “Navigating Narrow Spaces: A Comprehensive Framework for Agricultural Robots,” is available in IEEE Robotics and Automation Letters (DOI: 10.1109/LRA.2025.3592100).