BonnBot-I Plus: A Bio-diversity Aware Precise Weed Management Robotic Platform

TL;DR

The paper tackles the challenge of reducing herbicide use while preserving biodiversity by enhancing a precision weeding robot with biodiversity-aware planning. It introduces BonnBot-I Plus, featuring a rolling-view observation model, a bio-diversity-aware plant-level treatment scheme, and multi-nozzle intervention planning, validated in real sugar-beet fields and via the SB21 dataset. Key contributions include a 3.4–3.5% improvement in weeding performance over prior methods, a first demonstration of biodiversity-conscious weed management, and detailed real-world metrics showing planning and vision-system limitations contributing to losses of about $11.6\%$ and $14.7\%$ respectively. The work demonstrates practical viability and outlines concrete improvements in perception, planning, and ecological integration to advance field-deployable, eco-friendly robotic weed management.

Abstract

In this article, we focus on the critical tasks of plant protection in arable farms, addressing a modern challenge in agriculture: integrating ecological considerations into the operational strategy of precision weeding robots like \bbot. This article presents the recent advancements in weed management algorithms and the real-world performance of \bbot\ at the University of Bonn's Klein-Altendorf campus. We present a novel Rolling-view observation model for the BonnBot-Is weed monitoring section which leads to an average absolute weeding performance enhancement of $3.4\%$. Furthermore, for the first time, we show how precision weeding robots could consider bio-diversity-aware concerns in challenging weeding scenarios. We carried out comprehensive weeding experiments in sugar-beet fields, covering both weed-only and mixed crop-weed situations, and introduced a new dataset compatible with precision weeding. Our real-field experiments revealed that our weeding approach is capable of handling diverse weed distributions, with a minimal loss of only $11.66\%$ attributable to intervention planning and $14.7\%$ to vision system limitations highlighting required improvements of the vision system.

Figures (7)

• Figure 1: An image from underneath the BonnBot-I Platform in sugar beet fields at Campus Klein Altendorf of the University of Bonn, prior to conducting Precision weed management. (n1-n4: Spray nozzles in purple, L1-L4: Linear axes in light-green, and the detection area of the front camera in cyan).
• Figure 2: The software architecture, including sensors (purple), vision perception, localization of robot base frame $\mathcal{F}_R$ (light blue), intervention planning, intervention planner, and weeding axes controllers (orange).
• Figure 3: Segment-view vs Rolling-view Planning; two separate segments $T_0$ and $T_1$ with different weeding distributions are shown. An intermediate $T_i$ is substantially helping the planner to optimize the planned route of the weeding axis (green) w.r.t the baseline (red).
• Figure 4: Biodiversity in Focus: A field area with low crop density, susceptible to invasive weeds, and the potential role of Dicot plants in controlling weed growth.
• Figure 5: Example image of dataset SB21 (left) and multi-class annotations representing different types of crops and weeds using different colors (right).