Design and experiments with a SLAM system for low-density canopy environments in greenhouses based on an improved...

Area of Science:

• Robotics
• Agricultural Engineering
• Computer Vision

Background:

• Simultaneous Localization and Mapping (SLAM) algorithms struggle with accuracy in low-density greenhouse canopies.
• Robust navigation for agricultural robots is crucial for efficient crop management.

Purpose of the Study:

• To develop and evaluate a novel SLAM method for precise mapping and localization in greenhouse environments with sparse vegetation.
• To improve the performance of agricultural robots operating autonomously within crop canopies.

Main Methods:

• A multiline LiDAR-based SLAM approach was developed, integrating spatial downsampling and an adaptive filtering point cloud projection (AF-PCP) algorithm.
• The method utilizes wheel odometry and 16-line LiDAR data with adaptive vertical projections for map construction and pose estimation.
• Experiments were conducted on suspended tomato plants with varying leaf area densities (LADs).

Main Results:

• The AF-PCP SLAM algorithm significantly increased the average mapping area of crop rows by 155.7% compared to the standard Cartographer algorithm.
• Mean error and coefficient of variation for crop row length were reduced by 77.9% and 87.5%, respectively.
• Average relative localization errors were low (0.026–0.046 m) across different speeds, with a 79.9% reduction compared to track deduction algorithms.

Conclusions:

• The proposed AF-PCP SLAM framework enables precise mapping and localization for agricultural robots in low-density greenhouse canopies.
• The approach demonstrates robust performance and holds significant promise for autonomous navigation applications in precision agriculture.
• Enhanced mapping and localization capabilities are critical for the advancement of automated farming systems.