Adaptive LiDAR Odometry and Mapping for Autonomous Agricultural Mobile Robots in Unmanned Farms

TL;DR

This work tackles robust LiDAR-only localization and mapping for autonomous agricultural robots operating in unstructured fields with motion distortion and dynamic elements. It introduces AG-LOAM, a cascaded framework where dense GICP-based odometry is paired with an adaptive mapper that updates the map only when motion is stable and point correspondences are consistent, all without requiring IMU data. The key contributions include the two-filter adaptive mapping (Motion Stability Filter and Mapping Consistency Filter), extensive field evaluation across diverse planting and terrain conditions, ablation studies, and public release of code and datasets, with generalization validated on the TreeScope dataset. The framework delivers centimeter-scale odometry and robust mapping in real time, enabling reliable autonomous operation in agricultural environments and reducing dependence on external sensing modalities while offering avenues for future multi-modal extensions and fully autonomous field planning.

Abstract

Unmanned and intelligent agricultural systems are crucial for enhancing agricultural efficiency and for helping mitigate the effect of labor shortage. However, unlike urban environments, agricultural fields impose distinct and unique challenges on autonomous robotic systems, such as the unstructured and dynamic nature of the environment, the rough and uneven terrain, and the resulting non-smooth robot motion. To address these challenges, this work introduces an adaptive LiDAR odometry and mapping framework tailored for autonomous agricultural mobile robots operating in complex agricultural environments. The proposed framework consists of a robust LiDAR odometry algorithm based on dense Generalized-ICP scan matching, and an adaptive mapping module that considers motion stability and point cloud consistency for selective map updates. The key design principle of this framework is to prioritize the incremental consistency of the map by rejecting motion-distorted points and sparse dynamic objects, which in turn leads to high accuracy in odometry estimated from scan matching against the map. The effectiveness of the proposed method is validated via extensive evaluation against state-of-the-art methods on field datasets collected in real-world agricultural environments featuring various planting types, terrain types, and robot motion profiles. Results demonstrate that our method can achieve accurate odometry estimation and mapping results consistently and robustly across diverse agricultural settings, whereas other methods are sensitive to abrupt robot motion and accumulated drift in unstructured environments. Further, the computational efficiency of our method is competitive compared with other methods. The source code of the developed method and the associated field dataset are publicly available at https://github.com/UCR-Robotics/AG-LOAM.

Figures (15)

• Figure 1: Assorted views from the agricultural fields at the University of California, Riverside, where experiments were conducted. (a) A sample field featuring flat ground and navel orange trees planted in uniform, squared-shape pattern. (b) A sample field featuring a furrow on the side and a mixture of citrus trees planted in-row, with some space in between. (c) The Clearpath Robotics Jackal mobile robot used in this work shown while navigating through a field of densely in-row planted citrus trees. (d) The Jackal mobile robot captured at an instance of traversing a furrow while navigating in the field.
• Figure 2: Overall system diagram of the adaptive LiDAR odometry and mapping framework developed in this work. Solid arrows indicate the flow of point cloud data, and dashed arrows indicate information exchange between modules.
• Figure 3: An illustration of the operation principle of the Motion Stability Filter. In this example, map updates are permitted when the robot moves smoothly from $\mathbf{T}_{k-1}$ to $\mathbf{T}_{k}$. However, updates are rejected when the robot moves from $\mathbf{T}_{k}$ to $\mathbf{T}_{k+1}$, where an abrupt motion change is detected.
• Figure 4: An illustration of the operation principle of the Mapping Consistency Filter. In this example, points on the left and top line segments are selected because they are in close proximity to either the previous or the next frame. However, points on the right line segment are considered inconsistent and are rejected, as they only appear once in the current frame.
• Figure 5: The Clearpath Robotics Jackal mobile robot used in this work is shown at its departure (initial) position in the field. (a) In the early-stage design of our sensor payload, only the Velodyne VLP-16 LiDAR sensor was used for data collection (sequences A1-A6 and B1-B5). (b) The latest sensor payload design included LiDAR and GPS-RTK sensors for data collection (sequences C1-C7).