Towards Over-Canopy Autonomous Navigation: Crop-Agnostic LiDAR-Based Crop-Row Detection in Arable Fields

TL;DR

This work tackles the problem of autonomous navigation in row-crop fields without reliance on RTK-GPS by introducing a LiDAR-based over-canopy crop-row detection method that remains robust across crop types, growth stages, weeds, and canopy-occluded inter-row spaces. The approach combines a three-stage pipeline—crop-row detection, nonlinear MPC-based following, and lane-switching—to enable crop-agnostic field coverage, validated in both Gazebo simulations and real crops, achieving a cross-track error of about $3.55\,\text{cm}$ and accurate row-line predictions of approximately $3.35\,\text{cm}$ with $1.76^{\circ}$ orientation error. Key innovations include ground-plane-based LiDAR preprocessing, K-means clustering of row centroids, RANSAC-based 2D row line estimation, and a tightly integrated MPC/PID control framework. The results demonstrate practical viability for autonomous, GPS-independent navigation in diverse agricultural settings, with open-source implementations to accelerate adoption and further research.

Abstract

Autonomous navigation is crucial for various robotics applications in agriculture. However, many existing methods depend on RTK-GPS devices, which can be susceptible to loss of radio signal or intermittent reception of corrections from the internet. Consequently, research has increasingly focused on using RGB cameras for crop-row detection, though challenges persist when dealing with grown plants. This paper introduces a LiDAR-based navigation system that can achieve crop-agnostic over-canopy autonomous navigation in row-crop fields, even when the canopy fully blocks the inter-row spacing. Our algorithm can detect crop rows across diverse scenarios, encompassing various crop types, growth stages, the presence of weeds, curved rows, and discontinuities. Without utilizing a global localization method (i.e., based on GPS), our navigation system can perform autonomous navigation in these challenging scenarios, detect the end of the crop rows, and navigate to the next crop row autonomously, providing a crop-agnostic approach to navigate an entire field. The proposed navigation system has undergone tests in various simulated and real agricultural fields, achieving an average cross-track error of 3.55cm without human intervention. The system has been deployed on a customized UGV robot, which can be reconfigured depending on the field conditions.

Figures (9)

• Figure 1: Amiga robot navigating within soybean fields (left) and robot configuration (right). Our navigation system is a LiDAR-based autonomous navigation system for over-canopy navigation in row-crop fields.
• Figure 2: Navigation system's workflow. (i) The crop row detection algorithm uses LiDAR data and filtered odometry values $[x, y, \psi]$ as inputs, predicting crop rows in the form $[x1, y1, x2, y2]$ within the robot's frame. (ii) The crop row following algorithm applies nonlinear MPC to control the robot to follow the center line of the predicted rows, sending linear velocity $v$ and angular velocity $w$ commands. (iii) The crop row switching algorithm utilizes a PID controller to navigate the robot to the next lane if no more crop rows are detected.
• Figure 3: LiDAR point clouds filtering techniques. The virtual ground plane is estimated based on the LiDAR tilted angle $\theta$ and intersected with the centroids of the filtered LiDAR data. Points below the plane (blue) are removed, resulting in simplified LiDAR data (red) for crop-row detection.
• Figure 4: Illustration of the RANSAC line fitting algorithm for detecting crop rows. Centroids of the first row on the left and right in the robot's frame are extracted (left). The RANSAC line fitting algorithm is then applied (right) between the current farthest detected centroids and previous centroids within a specified range (e.g., 0.5 meters behind the robot). The predicted crop row is in the form of $[x_1, y_1, x_2, y_2]$.
• Figure 5: Gazebo simulated fields with three challenging scenarios ((i) Weeds and Discontinuities, (ii) Fully Blocked Canopies, and (iii) Curved rows.