UltraLight Luma: A Novel Edge-Deployable Perception Network for Crop-Row Segmentation in Agricultural Robotics

2026-10-08

AI Summary

This study addresses the challenge of crop row perception for agricultural robots operating under strict computational, memory, and power constraints by proposing UltraLight Luma, a lightweight segmentation network. The method employs a compact encoder-decoder architecture integrated with a progressive learning strategy and a visual servoing post-processing algorithm to achieve efficient crop row detection and navigation line extraction. With only 13.1k trainable parameters and a single-inference energy consumption as low as 25.23 mJ, the proposed model significantly outperforms U-Net and YOLO-based baselines. By striking an optimal balance between extreme lightweight design and high reliability, this work provides a resource-efficient solution for deploying robust crop row perception on edge devices in precision agriculture.

Abstract

Reliable crop-row perception is essential for autonomous agricultural robots, but low-cost deployment is constrained by computation, memory, power, and inference time. This paper addresses crop-row detection and navigation-line extraction using UltraLight Luma, a compact encoder-decoder segmentation network with a progressive learning strategy and only 13.1k trainable parameters. The predicted crop-row masks are processed using a post-processing algorithm to estimate row alignment for downstream visual servoing. UltraLight Luma improves parameter efficiency over U-Net, YOLOv8, and YOLOv26 while maintaining reliable crop-row detection performance. The model required only 25.23 mJ per inference, demonstrating its suitability for resource-constrained agricultural robots.

Problem

• crop-row segmentation
• agricultural robotics
• edge deployment
• resource-constrained perception

Innovation

• Edge-deployable perception network
• Crop-row segmentation
• Ultra-lightweight encoder-decoder
• Progressive learning strategy
• Parameter efficiency