LiteWeed: An Affordable Edge-AI Robotic Platform for Laser Weed Management in Resource-Constrained Environments

This study introduces LiteWeed, a compact and affordable semi-autonomous weed-management robot designed for operation in unstructured, resource-constrained agricultural environments. Unlike conventional systems that rely on high-cost infrastructure suitable only for structured fields, this system addresses the specific challenges of smallholder farms. It employs a lightweight, AI-driven YOLOX-nano architecture optimized for edge vision computing to detect and classify weeds, achieving a mean Average Precision (mAP@0.5:0.95) of 0.32 and a detection latency of 118 ms on a Raspberry Pi 5. To handle environmental variability such as uneven terrain and dynamic lighting, the robot features a “stop-and-align” control strategy coupled with a two-degree-of-freedom arm. Uniquely, laser exposure duration is dynamically tuned based on weed species and size to minimize energy consumption and ensure operational safety. Field trials in Vancouver, Canada, and Arusha, Tanzania, demonstrated the robot’s effectiveness, achieving weed-removal success rates of 97% and 96%, respectively, with a maximum laser exposure time of 60 seconds when targeting pigweed, purslane, and nutsedge. Designed to be cost-efficient and scalable (≈ $500 CAD), this integrated system offers a low-cost platform for early-stage, low-density, localized laser spot treatment, with the potential to reduce herbicide use and enhance weed-targeting precision for small farms.