In July 2023, we took advantage of two other ESA activities,  ALPER and RAPID, taking place in Bardenas Reales. By coordinating efforts, we were able to share programmatic and logistical resources, making the dataset collection process more efficient.

The Netherlands: Preparations in Motion

Before the tires of our rover touched down on Bardenas Reales, preparations were underway between ESA’s Planetary Robotics Lab (PRL) and the University of Málaga’s Space Robotics Lab (SRL) . The rover received a host of new upgrades, including a thermal camera, an enhanced IMU with GNSS capabilities, and force-torque sensors. Meanwhile, UMA’s department of Applied Physics I was responsible for preparing the portable spectrometer, which would enable the gathering of terrain composition data along the traverses.

We conducted tests with the GNSS and WiFi antennas to understand potential limitations in reach and visibility.  These tests informed our planning, allowing us to plan the placement for the WiFi repeater to reach the most reliable connectivity while covering the most terrain.

For this activity and this rover, we transitioned to using ROS 2, building on our prior experience with the Rock framework. While we had worked with other rover platforms and frameworks in the past, the integration of new software always brings its own set of challenges. The real test for this system integration would be out in the desert, where the combination of rough terrain, environmental factors, and extended operations would push both hardware and software to their limits.

One key part of the preparation involved scouting Bardenas Reales itself.  Colleagues who visited the site with industry partners of the parallel activities reported of a great terrain diversity ­— rocky outcrops, sandy patches, dried-out river beds — ideal for simulating planetary exploration. Their visit confirmed that this would be a great testbed for our rover and they identified a general area for the upcoming tests.

The van and mobile control station has a large payload battery installed and a solar panel on the roof to provide additional autonomy. Shortly before heading off to the field test, however, we noticed that the autonomy was worse than expected. Turns out that the solar panel setup was faulty and we scrambled to organize a portable solar panel for the test.

The Journey: From the Netherlands to Spain

To cover the 1,500 km journey from the Netherlands to Spain, two team members drove our fully equipped van — carrying the rover, supporting hardware, plus a few items for the parallel activities — over the course of two days. Meanwhile, the remaining two colleagues from the Netherlands flew in, while our Málaga-based team members traveled North by train.

Upon arrival, we surveyed the previously selected general test area to see how the vegetation had changed. Having gained a better understanding of our communication constraints, we focused on visibility into and out of the riverbed to identify optimal parking spots and locations for the WiFi repeater. Additionally, we prioritized sections of the terrain based on variety, traversability, potential coverage, and proximity to suitable van locations. Ideally, the van would maintain a connection to cellular networks while ensuring WiFi coverage over the rover’s key traverses. Balancing connectivity with terrain coverage was crucial, because frequent relocations and the setup of the base station and infrastructure would consume valuable time.

Into the Field

The Bardenas Reales semi-desert environment posed its own set of challenges.  Temperatures soared during the day, which was not only taxing for the hardware but also for the team. We had accounted for this by upgrading all 3D printed PLA parts to PET-G and by scheduling the recordings for the mornings, a longer break during the midday heat, and continued work into the evening, to keep ourselves and our systems cool enough to function reliably. This time would also serve to top up the van batteries by continuing solar charging but removing most or any loads. Tear-down, transit, break, and setup would take too long however, so we adapted the plan and worked throughout the entire day instead. Also, there was more wind than expected, which created some problems, but reduced the problems from the heat.

While we had anticipated some connectivity issues, things worked surprisingly well in the field. RTK corrections, which were crucial for centimeter-level accuracy in positioning, came through over the LTE link so we didn’t even use our dedicated back-up base station.

During data collection, we had to constantly ensure that all the sensors were functioning as expected. Monitoring and taking notes was a full-time task besides communicating with the rover operators, carrying batteries, moving the repeater and aiming the base WiFi antenna. For these reasons we were happy to have two people in the van and two with the rover at all times.

Post-Mission: Processing and Preparing for Publication

The weeks following the fieldwork were devoted to processing the collected data. Delogging — sorting through and cleaning up sensor data logs — was a time-consuming but vital part of the post-mission work (unprocessed sensor data is available, too, and so are our scripts). We identified and documented sensor offsets and time intervals for the highest quality subsets of traverse data for fellow researchers tackling similar planetary exploration problems. Despite the overall reliable RTK corrections, we did notice some instances where the GNSS would jump. Whenever we noticed this in the field, we stopped moving the rover plus we highlighted the affected sections of the trajectories in the traverse overviews.

Our experience and early findings were presented at ESA’s ASTRA 2023 conference while van specific details were discussed in a separate presentation at the same conference. This gave us an opportunity to discuss the challenges and lessons learned, from preparing such a dataset collection to navigating rough terrain. The feedback and insights from that conference session informed the documentation of the dataset and helped us, and hopefully others, refine our data collection techniques for future missions.

Beyond BASEPROD

With the BASEPROD dataset now publicly available, we’re excited to see how the community will use it. The dataset covers just short of two kilometers of traverses and provides information that can advance research on planetary locomotion, terrain classification, and rover navigation.  We hope it will serve as a valuable resource, sparking new discoveries and innovations in the field of space exploration.

For more details, please check out the linked BASEPROD paper and explore the links to the involved institutions.