Field Campaign Update: LiDAR-Based Soil Surface Roughness Assessment with FarmDroid FD20

As part of our ongoing research into robot–soil interactions, the SoilRob team conducted a field trial on July 1–3, 2026, at ZALF in Müncheberg. The campaign evaluated a LiDAR-based soil surface roughness scanning system mounted directly onto the autonomous FarmDroid FD20 mechanical weeding robot.

The primary goal of this experiment is to quantify how repeated autonomous robot passes during mechanical weeding alter the soil surface microtopography, while validating continuous LiDAR surface modeling against traditional reference methods.

Key Setup & Treatments

  • Autonomous Platform: A FarmDroid FD20 solar-powered field robot operated across a bare soil trial layout divided into longitudinal blocks and transverse sections.
  • Traffic Intensities: Twelve traffic lanes were established to test four distinct robot traffic intensities: Control (0 passes), Low (3 passes), Medium (7 passes), and High (14 passes).
  • Primary Sensor: A Pepperl+Fuchs R2000 2D LiDAR scanner ("Ackerscanner") was integrated onto a custom multisensor frame on the rear of the FD20, continuously scanning the freshly cultivated soil at 35 lines per second.

Reference & Supporting Soil Measurements

To evaluate and validate the LiDAR-derived metrics, comprehensive soil physical parameters were recorded across all experimental blocks:

  • Pinboard Roughness: High-resolution manual pinboard photographs were taken at 7.5 m intervals to extract Random Roughness (RR) benchmark values.
  • Soil Moisture: Volumetric soil moisture was monitored at designated interval points (ACCLIMA sensor) and continuously in control lanes (TOMST sensor).

Soil Structure & Hydraulics: Core sampling was performed for bulk density and soil texture analysis, accompanied by saturated hydraulic conductivity and infiltration capacity tests (METER SATURO).

Next Steps & Outlook

Data processing workflows are currently underway to generate detrended digital height models from the raw LiDAR point clouds. Key roughness parameters, including Root Mean Square height (RMS), Mean Absolute Deviation (𝑅𝑎), skewness, and Random Roughness (RR), are being extracted and cross-validated with pinboard data.

Go back