Resilient Terrain Navigation with a 5 DOF Metal Detector Drone

Patrick Pfreundschuh, Rik Bähnemann, Tim Kazik, Thomas Mantel, Roland Siegwart, Olov Andersson

Abstract

Micro aerial vehicles (MAVs) hold the potential for performing autonomous and contactless land surveys for the detection of landmines and explosive remnants of war (ERW). Metal detectors are the standard detection tool but must be operated close to and parallel to the terrain. A successful combination of MAVs with metal detectors has not been presented yet, as it requires advanced flight capabilities. To this end, we present an autonomous system to survey challenging undulated terrain using a metal detector mounted on a 5 degrees of freedom (DOF) MAV. Based on an online estimate of the terrain, our receding-horizon planner efficiently covers the area, aligning the detector to the surface while considering the kinematic and visibility constraints of the platform. As the survey requires resilient and accurate localization in diverse terrain, we also propose a factor graph-based online fusion of GNSS, IMU, and LiDAR measurements. We validate the robustness of the solution to individual sensor degeneracy by flying under the canopy of trees and over featureless fields. A simulated ablation study shows that the proposed planner reduces coverage duration and improves trajectory smoothness. Real-world flight experiments showcase autonomous mapping of buried metallic objects in undulated and obstructed terrain.

Introduction

Autonomous landmine detection is one of the most dangerous and technically demanding tasks in robotics. While drones have long been proposed as a solution for humanitarian demining, the physics of metal detection present a significant challenge: the sensor must be kept very close to the ground and perfectly parallel to the surface, even as the terrain rolls and dips. Traditional underactuated quadrotors struggle with this; to move forward, they must tilt their entire body, which moves a rigidly attached sensor away from the optimal detection angle. Furthermore, demining often occurs in "denied" environments, such as under tree canopies (GNSS-denied) or in featureless fields where LiDAR-based SLAM struggles to find anchors.

System Design and Methodology

The proposed system addresses these challenges through a specialized hardware platform and a resilient software stack built on three pillars.

5-DOF Hardware Platform

The researchers utilized the Voliro-T, a tricopter with tilting rotors. This 5-DOF configuration allows for independent control of position and tilt (pitch and heading), enabling the drone to "hug" uneven terrain while maintaining a constant height and sensor alignment without coupling translational movement to body orientation.

The 5-DOF MAV platform equipped with GNSS, IMU, LiDAR, and a metal detector.

Resilient State Estimation

To ensure reliable navigation across diverse terrains, a factor graph-based fusion system was implemented. This system performs real-time extrinsic calibration between local odometry and global GNSS frames. By treating sensor relationships as dynamic variables, the system maintains a consistent trajectory even when LiDAR "slips" in tall grass or GNSS signals are lost under dense foliage.

Receding-Horizon Planner

Instead of a reactive "greedy" approach, the system employs a receding-horizon planner that looks ahead several meters. This planner optimizes for smooth yaw and pitch trajectories, satisfying ground-alignment constraints while minimizing erratic, battery-draining motions. A key insight from the study is that allowing a small alignment error (within the sensor's 7.5° tolerance) significantly improves flight efficiency and smoothness compared to attempting perfect alignment at every millisecond.

Experimental Results

The system was validated through both simulation and real-world field tests at ETH Zurich, including dense tree canopies and featureless fields.

Localization and Efficiency

In simulations, the receding-horizon planner was 10.6% faster than standard baselines and reduced yaw changes by approximately 24° per step. In real-world tests involving total GNSS loss and LiDAR drift, the system maintained a landing accuracy of 0.15m, significantly outperforming standard LiDAR-Inertial Odometry (LIO), which deviated by over 1.5m.

Autonomous Mapping

The system successfully mapped buried metallic objects in complex, undulated environments. In one test, it identified 9 out of 9 buried gas cartridges. In a second test featuring physical obstacles, it successfully avoided a pole while mapping 8 out of 9 targets, only missing the target directly adjacent to the obstacle for safety reasons.

Autonomous mapping results showing the 3D terrain mesh, 2D detection heatmap, and the MAVs in the field.

Conclusion

This work demonstrates that a 5-DOF MAV combined with resilient state estimation and look-ahead planning can effectively perform autonomous landmine detection in challenging terrains. While the current obstacle avoidance is reactive, future work aims to integrate it directly into the coverage planner. Additionally, moving away from "black box" sensors toward those providing raw signal data could further improve detection sensitivity and consistency.

Summary updated

Abstract

Micro aerial vehicles (MAVs) hold the potential for performing autonomous and contactless land surveys for the detection of landmines and explosive remnants of war (ERW). Metal detectors are the standard detection tool but must be operated close to and parallel to the terrain. A successful combination of MAVs with metal detectors has not been presented yet, as it requires advanced flight capabilities. To this end, we present an autonomous system to survey challenging undulated terrain using a metal detector mounted on a 5 degrees of freedom (DOF) MAV. Based on an online estimate of the terrain, our receding-horizon planner efficiently covers the area, aligning the detector to the surface while considering the kinematic and visibility constraints of the platform. As the survey requires resilient and accurate localization in diverse terrain, we also propose a factor graph-based online fusion of GNSS, IMU, and LiDAR measurements. We validate the robustness of the solution to individual sensor degeneracy by flying under the canopy of trees and over featureless fields. A simulated ablation study shows that the proposed planner reduces coverage duration and improves trajectory smoothness. Real-world flight experiments showcase autonomous mapping of buried metallic objects in undulated and obstructed terrain.
BibTeX
@inproceedings{pfreundschuh2023resilient,
  author = {Patrick Pfreundschuh and Rik Bähnemann and Tim Kazik and Thomas Mantel and Roland Siegwart and Olov Andersson},
  title = {{Resilient Terrain Navigation with a 5 DOF Metal Detector Drone}},
  booktitle = {IEEE International Conference on Robotics and Automation},
  year = {2023},
  doi = {10.1109/icra48891.2023.10161253},
}
Figure