Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Autonomous Soil Collection in Environments With Heterogeneous Terrain

Andrew Dudash, Beyonce Andrades, Ryan Rubel, Mohammad Goli, Nathan Clark, William Ewald

TL;DR

This work tackles autonomous soil collection in heterogeneous amorphous terrain by deploying a general-purpose 7-DOF robotic arm equipped with a wrist camera and a two-finger gripper. A vision-guided pipeline uses semantic segmentation via a U-Net, followed by contour-based localization and 3D pose estimation to select soil regions, while MoveIt with an RRT* planner and a proportional depth controller execute pickup. The authors validate the prototype through arm-performance and vision-ablation experiments, reporting robust soil collection in both homogeneous and heterogeneous terrains and showing that data augmentation in training substantially improves segmentation. The results demonstrate the feasibility of autonomous soil sampling without niche hardware, highlighting practical implications for robotic agriculture, surveying, and hazard detection, and outlining directions for outdoor testing and control-system enhancements.

Abstract

To autonomously collect soil in uncultivated terrain, robotic arms must distinguish between different amorphous materials and submerge themselves into the correct material. We develop a prototype that collects soil in heterogeneous terrain. If mounted to a mobile robot, it can be used to perform soil collection and analysis without human intervention. Unique among soil sampling robots, we use a general-purpose robotic arm rather than a soil core sampler.

Autonomous Soil Collection in Environments With Heterogeneous Terrain

TL;DR

This work tackles autonomous soil collection in heterogeneous amorphous terrain by deploying a general-purpose 7-DOF robotic arm equipped with a wrist camera and a two-finger gripper. A vision-guided pipeline uses semantic segmentation via a U-Net, followed by contour-based localization and 3D pose estimation to select soil regions, while MoveIt with an RRT* planner and a proportional depth controller execute pickup. The authors validate the prototype through arm-performance and vision-ablation experiments, reporting robust soil collection in both homogeneous and heterogeneous terrains and showing that data augmentation in training substantially improves segmentation. The results demonstrate the feasibility of autonomous soil sampling without niche hardware, highlighting practical implications for robotic agriculture, surveying, and hazard detection, and outlining directions for outdoor testing and control-system enhancements.

Abstract

To autonomously collect soil in uncultivated terrain, robotic arms must distinguish between different amorphous materials and submerge themselves into the correct material. We develop a prototype that collects soil in heterogeneous terrain. If mounted to a mobile robot, it can be used to perform soil collection and analysis without human intervention. Unique among soil sampling robots, we use a general-purpose robotic arm rather than a soil core sampler.
Paper Structure (6 sections, 8 figures, 4 tables)

This paper contains 6 sections, 8 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. ARCHITECTURE
  4. EXPERIMENTS
  5. RESULTS
  6. CONCLUSION

Figures (8)

  • Figure 1: The prototype described in this system discovers and collects soil samples in the situation shown in Figure \ref{['fig:hetero-material']}.
  • Figure 2: Our prototype uses an arm with 7-DOF, a wrist mounted camera, and a two finger end effector.
  • Figure 3: We use a segmentation model to classify each pixel of the image and then we use a border following algorithm to break the segmented image into groups of pickable objects.
  • Figure 4: The distance detected by the infrared sensor is used as feedback.
  • Figure 5: For some stages, the robot can retry on failure.
  • ...and 3 more figures