Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Eyes on the Grass: Biodiversity-Increasing Robotic Mowing Using Deep Visual Embeddings

Lars Beckers, Arno Waes, Aaron Van Campenhout, Toon Goedemé

TL;DR

The paper tackles biodiversity loss in urban lawns by turning robotic mowing into an active ecological intervention. It proposes a vision-based framework that uses a PlantNet300K-pretrained ResNet50 to embed vegetation in a feature space and estimate biodiversity via global dispersion and local density. A two-stage mowing procedure—patrol for embedding space construction and biodiversity-aware mowing decision—drives selective mowing that preserves diverse patches. Experiments on mock-up and real lawns show embedding-space dispersion correlates with expert biodiversity scores and demonstrate feasibility for real-world deployment to transform lawns into biodiverse biotopes.

Abstract

This paper presents a robotic mowing framework that actively enhances garden biodiversity through visual perception and adaptive decision-making. Unlike passive rewilding approaches, the proposed system uses deep feature-space analysis to identify and preserve visually diverse vegetation patches in camera images by selectively deactivating the mower blades. A ResNet50 network pretrained on PlantNet300K provides ecologically meaningful embeddings, from which a global deviation metric estimates biodiversity without species-level supervision. These estimates drive a selective mowing algorithm that dynamically alternates between mowing and conservation behavior. The system was implemented on a modified commercial robotic mower and validated both in a controlled mock-up lawn and on real garden datasets. Results demonstrate a strong correlation between embedding-space dispersion and expert biodiversity assessment, confirming the feasibility of deep visual diversity as a proxy for ecological richness and the effectiveness of the proposed mowing decision approach. Widespread adoption of such systems will turn ecologically worthless, monocultural lawns into vibrant, valuable biotopes that boost urban biodiversity.

Eyes on the Grass: Biodiversity-Increasing Robotic Mowing Using Deep Visual Embeddings

TL;DR

The paper tackles biodiversity loss in urban lawns by turning robotic mowing into an active ecological intervention. It proposes a vision-based framework that uses a PlantNet300K-pretrained ResNet50 to embed vegetation in a feature space and estimate biodiversity via global dispersion and local density. A two-stage mowing procedure—patrol for embedding space construction and biodiversity-aware mowing decision—drives selective mowing that preserves diverse patches. Experiments on mock-up and real lawns show embedding-space dispersion correlates with expert biodiversity scores and demonstrate feasibility for real-world deployment to transform lawns into biodiverse biotopes.

Abstract

This paper presents a robotic mowing framework that actively enhances garden biodiversity through visual perception and adaptive decision-making. Unlike passive rewilding approaches, the proposed system uses deep feature-space analysis to identify and preserve visually diverse vegetation patches in camera images by selectively deactivating the mower blades. A ResNet50 network pretrained on PlantNet300K provides ecologically meaningful embeddings, from which a global deviation metric estimates biodiversity without species-level supervision. These estimates drive a selective mowing algorithm that dynamically alternates between mowing and conservation behavior. The system was implemented on a modified commercial robotic mower and validated both in a controlled mock-up lawn and on real garden datasets. Results demonstrate a strong correlation between embedding-space dispersion and expert biodiversity assessment, confirming the feasibility of deep visual diversity as a proxy for ecological richness and the effectiveness of the proposed mowing decision approach. Widespread adoption of such systems will turn ecologically worthless, monocultural lawns into vibrant, valuable biotopes that boost urban biodiversity.

Paper Structure

This paper contains 16 sections, 1 equation, 6 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. METHODOLOGY
  4. Robot Demonstrator
  5. Proposed Mowing Procedure
  6. Stage 1: Exploratory Patrol and Image Acquisition
  7. Stage 2: Biodiversity-Aware Mowing Procedure
  8. EXPERIMENTS & RESULTS
  9. Datasets
  10. Biodiversity estimation with deep embeddings
  11. Choice of the pre-training dataset
  12. Visualizing the Embedding Space
  13. Biodiversity-Aware Mowing Experiments
  14. Mock-up Lawn Experiment
  15. Evaluation on Real Lawn Data
  16. ...and 1 more sections

Figures (6)

  • Figure 1: Adapted robotic mower used in the experiments. The camera observes the vegetation in front of the blades; the onboard controller switches the cutting blades on or off based on biodiversity estimation.
  • Figure 2: Overview of the mowing procedure
  • Figure 3: Example frames from the collected lawn datasets, illustrating variation in biodiversity and mowing conditions. (a)-(d): datasets with large biodiversity. (e)-(g): datasets with less biodiversity. (h): Image taken with our robot during the mock-up lawn experiment. These datasets are publicly available at https://github.com/Lars-Beckers/BioBot.
  • Figure 4: t-SNE visualization of embeddings from all biodiverse (red, blue) and non-biodiverse (orange, cyan) datasets using the PlantNet300K-pretrained ResNet50 model. Greater spatial spread indicates higher visual diversity within the dataset.
  • Figure 5: Mock-up lawn experiment. A video can be seen at https://www.youtube.com/shorts/n4Pn7bTxB2s
  • ...and 1 more figures