Navigation – Plan du site

AccueilNuméros45-12. Methods and innovationsArchGeoRobot: Automated Archaeo-G...

2. Methods and innovations

ArchGeoRobot: Automated Archaeo-Geophysical Data Acquisition Using an Unmanned Ground Vehicle

Lieven Verdonck
p. 219-221

Résumé

– An unmanned ground vehicle was used to perform an automated GPR survey.

– The robot pulling the GPR surveyed the area without human intervention.

– It remained close to the predefined path, usually with an accuracy better than 5 cm.

Haut de page

Texte intégral

1Large progress has been made in geophysical data acquisition and instrumentation. Vehicle-towed platforms, multi-channel arrays, and real-time positioning allow for the survey of large areas at a high spatial resolution. However, data acquisition remains time-consuming for the operator pulling the sensor, pushing the cart, or driving the vehicle. In order to automate geophysical data collection, one possibility is the use of unmanned aerial vehicles (UAVs).

2Several UAV-enabled magnetometers are already commercially available, while tests with drone-borne electromagnetic survey have been conducted (Karoulis et al., 2020). For other geophysical methods such as earth resistance survey, the use of UAVs seems less obvious because the geophysical sensor needs to be in contact with (or very close to) the surface, and because of the restricted payload of UAVs. Although the potential of drone-borne GPR measurements has been demonstrated (Wu et al., 2019), field tests at archaeological sites showed a reduced penetration depth of air-coupled GPR arrays compared to ground-coupled systems (Linford et al., 2015). Therefore also ground-based methods for automating geophysical data acquisition should be explored.

3In what follows, a small unmanned ground vehicle (UGV) for automated GPR prospection is presented (Fig. 1). UGVs are used for example in precision agriculture. In combination with GPR, they are used for the monitoring of ice sheets, the inspection of infrastructure such as bridges (Gibb et al., 2018), and humanitarian demining. The UGV used in this project is a four-wheeled, differential drive mobile robot of ~1 m × 0.65 m size, with a maximum speed of 1 m/s and a payload of up to 75 kg.

Figure 1. The UGV and GPR used for the automated prospection experiment. An additional RTK GNSS receiver is mounted above the GPR antennae.

Figure 1. The UGV and GPR used for the automated prospection experiment. An additional RTK GNSS receiver is mounted above the GPR antennae.

4The UGV is equipped with several sensors. The wheel encoders measure the instantaneous wheel velocities and provide a position estimate. The robot’s inertial measurement unit records the acceleration, the angular velocity and the magnetic field strength, and provides an orientation estimate. A GNSS receiving corrections from permanent reference stations via mobile internet allows an accurate estimation of the robot’s location. In combination with the GNSS, tracking total station measurements were used to compute the UGV’s heading in an accurate way (instead of the total station, a second GNSS receiver could be used). To solve the state estimation problem (position, orientation and velocity), sensor data are fused by means of an extended Kalman filter (EKF), consisting of a prediction step, which projects the current state estimate forward in time, and a correction step, which corrects the prediction using the sensor measurements (Moore & Stouch, 2016). The EKF algorithm is part of the Robot Operating System (ROS), an open-source robotic framework structured as a large number of small programs (‘nodes’) passing messages to one another (Quigley et al., 2016). A further requirement for automated GPR data acquisition is that the UGV can navigate autonomously. This is implemented by another ROS node, which determines a path towards a goal and sends velocity commands to the UGV.

5To demonstrate the use of a UGV for automated geophysical prospection, an experiment was conducted in which the robot pulled two 500 MHz GPR antennae over a flat asphalt surface of ~30 m × 16 m without any obstacles (Fig. 1). It was given a series of waypoint goals making up a rectangular path with rounded-off corners, gradually shifting to the south to cover the entire area. The first waypoints are shown in Figure 2b. The separation between the antennae (25 cm) was too large for an unaliased GPR survey (Verdonck et al., 2015). To reduce the transect spacing from 25 cm (the red lines in Fig. 2a) to 12.5 cm (red and green lines), the distance between passes alternated between 12.5 cm and 37.5 cm (Fig. 2a).

Figure 2. (a) data recording scheme: the even-numbered passes reduced the transect spacing from 25 cm to 12.5 cm. (b) The UGV covered the prospection area evenly. The first goals (of a series of 256 goals in total) are shown as red dots. (c) Detail of (b), showing a comparison between the measurements taken by the GNSS receiver mounted above the GPR antennae (in green), and the predefined path (in orange). Whereas the UGV deviated from the predefined path in the curved parts at the beginning of the lines, the error was small in the straight parts.

Figure 2. (a) data recording scheme: the even-numbered passes reduced the transect spacing from 25 cm to 12.5 cm. (b) The UGV covered the prospection area evenly. The first goals (of a series of 256 goals in total) are shown as red dots. (c) Detail of (b), showing a comparison between the measurements taken by the GNSS receiver mounted above the GPR antennae (in green), and the predefined path (in orange). Whereas the UGV deviated from the predefined path in the curved parts at the beginning of the lines, the error was small in the straight parts.

6The robot completed the survey in approximately 2.5 hours. The navigation nodes in ROS require a considerable amount of parameter tuning, e.g. the tolerance within which the robot should reach a goal. A small tolerance caused the UGV to get stuck during the survey since it could not reach its goal, requiring human intervention. After the tolerance had been increased, the robot covered the complete predefined route independently, although it interrupted the prospection a few times when the RTK GNSS signal was temporarily unavailable (a safety node then prevents the robot from moving). An important question is how accurately the robot followed the predefined lines. Coordinate precision for spatially unaliased GPR surveys (i.e. instrument precision and precision when following the grid lines) should be better than half the required transect spacing (Grasmueck & Viggiano, 2007). Often the required coordinate precision is less than 5 cm (Verdonck et al., 2015). From Figure 2b, it is clear that the robot covered the test area evenly. The root-mean-square error between predefined coordinates and the closest measured coordinate was ~2 cm in in-line direction, and ~6 cm in cross-line direction. Figure 2c shows that it deviated most from the predefined lines near the corners at the beginning of the lines, whereas in the straight parts, the error was usually below 5 cm.

7There was no interference from the UGV sensors (e.g. the GNSS mobile phone) producing noise in the GPR data, when the stack was large enough. A GPR time-slice resulting from the survey is shown in Figure 3.

Figure 3. GPR time-slice resulting from the automated survey, at a two-way travel time of 11–12 ns.

Figure 3. GPR time-slice resulting from the automated survey, at a two-way travel time of 11–12 ns.

8The scale of the data acquisition presented here is small. The UGV is not powerful enough to pull large arrays. However, more powerful UGVs exist, to which the described principles can be applied with little modification. Although large UGVs represent a considerable investment, it can be expected that more automated data acquisition will become part of geophysical practice in the near future. Besides for GPR prospection, UGVs could be used for other frequently used geophysical techniques in archaeology. Further work will include tests on more uneven surfaces, and in fields where the UGV should avoid obstacles such as trees.

Haut de page

Bibliographie

Gibb, S., La, H.M., Le, T., Nguyen, R., Schmid, R., Pham, H., 2018. Nondestructive evaluation sensor fusion with autonomous robotic system for civil infrastructure inspection. Journal of Field Robotics, 35: 988-1004.

Grasmueck, M., Viggiano, D.A., 2007. Integration of ground-penetrating radar and laser positioning sensors for real-time 3-D data fusion. IEEE Transactions on Geoscience and Remote Sensing, 45: 130-137.

Karoulis, M., Ritsema, I., Bremmer, C., De Kleine, M., 2020. Drone-Borne Electromagnetic (DREM) Surveying in The Netherlands. In 26th European Meeting of Environmental and Engineering Geophysics. DOI: https://doi.org/10.3997/2214-4609.202020032.

Linford, N., Linford, P., Payne, A., 2015. First results from a new ground-coupled multi- element GPR array. Archaeologia Polona, 53: 631-635.

Moore, T., Stouch, D., 2016. A generalized extended Kalman filter implementation for the Robot Operating System. In E. Menegatti, N. Michael, K. Berns, H. Yamaguchi (eds.), Proceedings of the 13th International Conference on Intelligent Autonomous Systems. Springer, Cham, 335-348.

Quigley, M., Gerkey, B., Smart, W.D., 2016. Programming robots with ROS. A practical introduction to the Robot Operating System. O’Reilly Media Inc., Sebastopol, CA.

Verdonck, L., Taelman, D., Vermeulen, F., Docter, R., 2015. The impact of spatial sampling and migration on the interpretation of complex archaeological ground-penetrating radar data. Archaeological Prospection, 22: 91-103.

Wu, K., Arambulo Rodriguez, G., Zajc, M., Jacqemin, E., Clément, M., De Coster, A., Lambot, S., 2019. A new drone-borne GPR for soil moisture mapping. Remote Sensing of Environment, 235: 111456.

Haut de page

Table des illustrations

Titre Figure 1. The UGV and GPR used for the automated prospection experiment. An additional RTK GNSS receiver is mounted above the GPR antennae.
URL http://journals.openedition.org/archeosciences/docannexe/image/9718/img-1.jpg
Fichier image/jpeg, 1,8M
Titre Figure 2. (a) data recording scheme: the even-numbered passes reduced the transect spacing from 25 cm to 12.5 cm. (b) The UGV covered the prospection area evenly. The first goals (of a series of 256 goals in total) are shown as red dots. (c) Detail of (b), showing a comparison between the measurements taken by the GNSS receiver mounted above the GPR antennae (in green), and the predefined path (in orange). Whereas the UGV deviated from the predefined path in the curved parts at the beginning of the lines, the error was small in the straight parts.
URL http://journals.openedition.org/archeosciences/docannexe/image/9718/img-2.jpg
Fichier image/jpeg, 4,2M
Titre Figure 3. GPR time-slice resulting from the automated survey, at a two-way travel time of 11–12 ns.
URL http://journals.openedition.org/archeosciences/docannexe/image/9718/img-3.jpg
Fichier image/jpeg, 1,1M
Haut de page

Pour citer cet article

Référence papier

Lieven Verdonck, « ArchGeoRobot: Automated Archaeo-Geophysical Data Acquisition Using an Unmanned Ground Vehicle »ArcheoSciences, 45-1 | 2021, 219-221.

Référence électronique

Lieven Verdonck, « ArchGeoRobot: Automated Archaeo-Geophysical Data Acquisition Using an Unmanned Ground Vehicle »ArcheoSciences [En ligne], 45-1 | 2021, mis en ligne le 16 août 2021, consulté le 11 novembre 2025. URL : http://journals.openedition.org/archeosciences/9718 ; DOI : https://doi.org/10.4000/archeosciences.9718

Haut de page

Droits d’auteur

CC-BY-NC-ND-4.0

Le texte seul est utilisable sous licence CC BY-NC-ND 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont susceptibles d’être soumis à des autorisations d’usage spécifiques.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search