Tenerife Field Trials 2012
Minas de San José
On the way in Martian like landscape on Tenerife
Mobile systems are among the most critical in all space missions. In future, such platforms will require more autonomy than today to enhance their science return and effectiveness compared to the budget invested in such complex assets. The typical scouting rover scenario on Mars foresees autonomous traverses over several hundred meters between pre-defined waypoints (targets to be reached as global mission goals) within periods of at least one Martian day (1 sol = 24h 39min). The risk of passing potentially interesting targets en-route without recognizing them is high, since downlink data rate is not sufficient to transmit all observed images or videos.
Autonomy therefore needs to take into account the search for such potentially interesting targets while the rover is travelling. This complex system of hardware and software should be able to autonomously make decisions regarding navigation, image capture and sample recovery. The EU FP7-SPACE Project ProViScout (2010-2012), brings together 12 European research institutions and companies, and aims to prototype the robotics vision building blocks required by such a future autonomous exploration system. The algorithms developed for scientific target selection and navigation will be tested during the Field Trials from 10th to 19th of September in the Martian like landscape of El-Teide National Park on Tenerife.
Mini-Poster download: MiniPosterHandoutPRoViScoutEnglGerman.pdf (PDF, 1.8MB)
Mini-Poster download: MiniPosterHandoutPRoViScoutEnglSpanish.pdf (PDF, 1.8MB)
This was the last official trial day, and about ¼ of the team were going to leave Tenerife today. So, this was the day to run a complete sequence including all HW and SW parts and to present the results to our EU reviewer and officer.
The set up procedure was similar to the days before. At about 9:30 the EU reviewer and officer arrived on site. Dave and Gerhard explained them the system and all its components.
At about 10:00 Matt executed a sequence of 2 navigation steps with more than 30m each. The VO worked all the time properly and gave for the first section nearly the same results as the mechanical odometry. Idris stopped quite close to the set position and captured a panorama. Unfortunately, Science Assessment didn’t detect any targets. Afterwards, a DEM was generated from the WAC images. Trying to go 5m further to get the artificial targets (Dobbin and Kathrin) inside the FoV, Idris directly headed towards a rock. Because of the fixed PTU, the blind spot surrounding Idris was larger and the rock close to Idris was not represented in the hazard map.
Ernst and Honza repositioned the webcam for the real time video feet.
Later there were some problems with the global rover coordinates being provided for the overseer (Matt) and the 3D-ROV (3D visualization by George). About 30 minutes later, the correct rover position was sent to Matt and George. Idris drove on 3m and captured a panorama. Science Assessment detected a target and Idris captured a RGB image (gathering some detail information as kind of reaction, 11:30).
Derek P. and Mark explained the necessity for having tools like hammer or drill on the rover for geological relevant analysis.
Because the execution of a whole sequence (including 2 DEMs, navigation, panorama and Science Assessment) didn’t work faultlessly, the reviewer and the officer were requested to go back to the van and watch the 3D-ROV Demo.
Later, Laurence demonstrated the aerobot flying close to the van and capturing some images.
All the activities on site were broadcasted live via a real time video feet provided by CTU. Below you find some screenshots.
The overseer screen (SciSys):
The visualization by Trasys:
The field view webcam:
In the evening the first part of the final review meeting took place, where the achievement of the project’s objectives and the project results were discussed.
Like most of the days before, the day started quite early in the morning without the set up team having had breakfast. The first sensation was Laurence driving the van – he did very well and it was the first practice for driving the long way back home.
People from Electrosat arrived on site in time, started to set up the equipment for the Broadband Satellite application and even helped Steve to connect the Proviscout network to the world wide web.
Alberto, Fred, Matt und Bernie went to the site for setting up the rest (gazebo etc.). Before finally starting the tests, one more press team arrived and Alberto was interviewed once more.
Afterwards Alberto did some navigation tests, whereas Idris chose once more some „complicated“ path. But this time, Alberto realized why Idris was doing so (which was the first premise to solve this problem). Because of the fixed tilt, it was not possible to see the closest, Idris surrounding area, which again enlarges the non-defined area. Because non-defined points were automatically said to be hazards, many non-existing obstacles were detected and avoided.
CTU had installed a webcam on site which could also be seen via their real video screen.
At 12:27 a whole sequence including navigation, vision and science assessment was executed for the first time! Unfortunately the science assessment didn’t detect any of the targets inside the captured images.
Because last night there was a power outage at the Parador, Idris’ batteries had not been completely recharged overnight. So the AU team tried to recharge them by their generator during lunch. Because this didn’t work as expected, the generator was mounted on the rover. Idris was ready for „take off“ at 15:30.
At about 16:20 a sequence together with the 3D-ROV of George was successfully tested. Alberto watched the drive in the 3D visualization and commented it via radio.
At the end of the afternoon one could see and feel the exhaustion caused by the last days. But some team members didn’t take it too seriously, like Derek L., sitting and reading a book now and then, while the others tried hard to do their jobs. But that’s the advantage of the well prepared.
At the end of the test day (17:20), once more a whole sequence was executed and the natural "spectral" target could be captured. Due to too many shadows, the science assessment didn’t detect anything. So the sequence would be repeated the next day.
At 20:30 the official dinner took place at Tasca de Roberto in Vilaflor. For 20,- per person a delicious menu was served including salad, vegetables, potatoes, meat, seafood and dessert. Once more Conny was a master of organization and support. Steve, who decided to walk back to the Parador in the afternoon nearly had missed the dinner because of being late.
This day was Press day, which again ment getting up early for the set up team (AU plus JR). Most of the press were invited for between 16:00 and 18:00. Only Euronews, producing some documentation about astrobiology, was going to be on site in the morning.
When people entered the meeting room at 7:30, which was the time that was said to leave, the PTU fix of the evening before didn’t work anymore. So the broken tilt fixing unit seemed to be a real problem. Steve started repairing again (”it’s gonna be a long thing this time”).
So the rest started to have breakfast, waiting for further instructions by Gerhard.
The results of CTU’s DEM processing were the best news of the morning. The generated aerial DEM (based on the aerobot’s images captured one day before) could serve as initial DEM for Idris navigation.
Alberto and Matt stayed at the Parador for going on integrating the navigation SW to the executive, while Kathrin and Bernie went with Mark and Andy to the site taking Indie and the AU equipment (except Idris) and started to set up.
When arriving at the site, Euronews were already there.
At about 9:30 Laurence finally came and soon after Idris was brought with a “fixated” PTU that should be tested on site. Idris was already “labelled” by participating organizations’ labels. It looked quite good ;).
On her way for picking up the generator (for the Broadband Satellite application on 16th / 17th) and buying food and drinks for the day, Conny brought Kathrin to the airport (leaving test site at 10:30).
Back on site, Euronews were still there interviewing Gerhard with Idris in the background. Thus, Steve couldn’t even try his „Frankenstein“ PTU.
Because Bernie was free at the moment, he took over Kathrin’s job as „Official PRoViScout Photographer". People would almost have complained about all the releasing noise.
After Euronews had finished, it was Steve’s turn. The original euphoria soon gave way to disillusion: the repaired PTU couldn’t carry the weight and broke once more. So Steve fixed the tilt axis to a reasonable angle and Laurence adopted his software. Thus, people could go on working and testing.
Once more things didn’t go so well. Thus, Alberto wanted to reach a destination in about 2m distance, but Idris autonomously drove estimated 15m in a quite complicated curved way – finally reaching the expected goal.
Unfortunately the wind was very strong, so people had to hold the gazebo to prevent it from flying away.
At about 14:30 the wind became so strong, that the team removed the gazebo and went back to the van.
Lunch was already waiting (this time in SciSys‘ mini-van).
Soon after lunch first TV and radio teams arrived. All in all there was much more press and tourists than expected. Idris and Indie did their job very well and drove remote near the parking area being shot all the time. Especially families with children liked cute little Indie. Some of them even wanted to feed it!
Alberto, the only native Spanish speaker of our team, was giving interviews all the time.
Gerhard was interviewed several times in English and once even in Spanish!
Just in time for press Stefano D’Orilia, our EU officer, arrived on site.
After press had gone at about 18:00, we wanted to test a whole sequence, but unfortunately there were several bugs in Matt's Python-Scripts. So we packed up and left the site.
This day began very early again. The AU team plus Kathrin and Bernie left the Parador already at 7:30 to set up everything on test site; especially the aerobot was set up by Laurence and Steve. The rest came quite soon to see the aerobot fly.
Before, Bernie and Kathrin went along the whole trajectory to make sure, all landmarks being located at their original positions when areobot is capturing the test site.
The aerobot was ready to launch at 9:45. Fred and Steve navigated Idris and the aerobot along the whole traverse to capture images for 3D reconstruction. On its way back it flew several meters higher above the ground than it did in the first direction in order to cover the whole area by less images (to be processed later by CTU for 3D reconstruction). Every four seconds an image was captured (5MPx resolution, FoV 60degrees). Even though it would have been possible to steadily transfer the data via WiFi, the images were stored locally on the aerobot’s hard drive to make sure nothing being lost due to WiFi leaks.
At about 9:30 Conny brought breakfast to the site for those people not having had breakfast in the morning.
Bernie and Matt were sitting in the van trying to integrate the vision SW and the executive.
Alberto stayed at the Parador trying to improve his navigation SW. At about 10:00 he arrived on the site for updating and further testing on Idris.
At the same time Laurence, Steve and Fred brought the aerobot back to the van for having breakfast and transferring the captured image data. The images are very nice and Laurence (as the rest of us) was happy.
During midday, the SciSys team brought Indie to the site. Mark and Andy did some testing of their own vision and navigation SW.
Conny served lunch at about 13:00.
In the afternoon Matt, Alberto and Bernie managed to integrate the vision SW, the navigation SW and the science assessment to the executive. Idris went 20m autonomously! It seemed things are really going on, although there are still problems in detail to be solved. The figure below shows a navigation track (red) in dry-run mode with continuous guidance.
Testing science assessment once more was the last item of the day. Idris was manually driven to scientifically WP1 where “Kathrin”, “Dobbin” and the natural “spectral target” were located. AUPE and the two HyperCams took several images which were immediately processed. The results served as input for science assessment. Mainly the natural target gave promising results.
That evening the tests on site finished very late again. So the AU team, always being the first and the last on site, were quite tired. They were mainly pensive due to the PTU had been broken on its way back to the hotel. Steve was trying hard to repair and managed at about 21:00.
The meeting for planning the next day took place from 19:00 to about 20:00. One major topic was the press event on the next afternoon with at least 3 camera teams on site! Due to many nationalities involved in PRoViScout, it was easy for Gerhard to find native speakers in different languages being responsible for answering the questions of press and tourists’.
Michal and Honza proudly presented a DEM of the first third of the whole trajectory (calculated from the circular handheld images they had captured). Overnight the “coarse” aerobot data shall be processed as well – hopefully the amount of images and the level of detail are sufficient. Unfortunately there's not enough time to process the "fine" image series.
The schedule planning for the next day contained the following dialogue: Gerhard: “Perhaps the navigation thing will need a little bit longer.” Alberto: “Here in between you may give the rover to someone else - I'll have to fix the next bug.”
After dinner Conny was presenting some places on Fuerteventura that might be of interest for future projects.
Andrew was leaving today (hope the HyperCam images are ok).
At about 23:00 last people left for their rooms.
Steve, Fred and Mark left the Parador at 7:30 for early being on site and setting up Idris and the network. Soon after, Derek P. and DLR geologists as well as Andrew arrived to “manually” capture AUPE images for panorama and DEM processing as well as HyperCam images near a coarse way point close to the manually placed targets. Before, calibration took place.
At that time Bernie and Alberto stayed at the Parador’s meeting room for bug fixing. Laurence started setting up the aerobot.
Right before lunch at about 13:00 Bernie and Gerhard were smiling because of one more fixed VO bug.
That day Conny brought some fantastic food for lunch – Paella! Buffet was served in front of the van.
Right after lunch Bernie and Alberto went to Idris on site for trying the updated vision SW and testing autonomous navigation again.
When the DLR team came back from a walk, they asked what’s going on in the field. Fred shortly summarized: “They had a discussion about whether it’s an obstacle or not!” (16:17).
Only minutes later, a bus of Russian tourists arrived right next to the van. Several of them found the way to the test site. Gerhard was explaining what’s going on during the trials and allowed them to get closer to the rover. Idris was shot hundreds of times… very patient.
Bernie and Alberto were still testing VO and navigation. Mainly a set of parameters needed to be adapted, but Idris was already driving several meters autonomously without touching any obstacles. There still remained some problems to be solved before connecting the vision and navigation SW to the executive.
Meanwhile Laurence went on setting up the aerobot, so it was ready for take of in the late afternoon and the helium should arrive at 17:30. Gerhard decided to bring it onto the test site for a short initial drive in the evening. But due to PVS (the processing unit on Idris) was KIA (Killed In Action :) at about 17:00, we were packing up and left. Within an hour the team arrived in the hotel and people went on bug fixing and improving their SW and interfaces. PVS was repaired immediately by simply plugging in the hard drive connector.
Michal and Honza from CTU went on processing all circle images they captured during the last days. They also tried to “merge” the single resulting DEMs to one overall DEM, but they recognized some trouble as well. So everything seemed to be as it should be on such a field trial..
At about the same time Conny brought Mark to the airport. Bye, bye Mark!
Before dinner (19:30), there was an informal “get together” in the meeting room where people discussed the plans for the next day.
At 21:50 Dave (AU) arrived at his hotel in Vilaflor. George from TraSys also arrived in the evening at the Parador.
After breakfast at 8:00, SciSys started setting up their little rover Indie in the meeting room. It serves as backup for Idris in case of any problems and collects image data representative for a small rover mission.
Due to again having problems with the van’s vacuum pump (which was already “repaired” on the way to Tenerife), Idris could leave the Parador not until 9:15. About 1 hour later everything was set up on site.
Michal and Honza were the first on site. Michal went on capturing hand held images for the DEM creation, while Honza was sitting in the van (with power supply) and processed the images step by step. The idea was to generate 17 single DEMs, whereas each of these DEMs covers one circle containing five landmarks. The belonging (70-90) images were taken from outside the circle pointing to its centre direction. The single DEMs would finally be stitched by the known geo-coordinates of those landmarks visible in neighboured DEMs.
At about 10:00, Derek (ULEIC) arrived on site and brought four artificial targets to be detected by science assessment. He went to Parador again to meet people from SciSys.
The dGPS mounted on Idris served as reference for time synchronization of all cameras and devices. By knowing the offset to the visual system, AUPE’s exact position was logged any time dGPS was running.
Bernie, Alberto and Laurence tried again to start the vision SW. First Idris-DEMs were successfully computed. Unfortunately, the visual odometry appeared to be buggy as well.
At about 11:30 Steve installed some kind of safety tape to fulfil the requirements of our permit and to keep tourists away from the test site.
At about 13:30 a loose screw on PTU was fixed by Mark and Steve.
At 13:45 the SciSys team came to the site.
Together with Ernst and Derek P. they had a look at the planned trajectory and pointed out an optimal position for placing the artificial targets. This location was settled to be the first coarse waypoint (WP). The targets were named Obsidian, Dobbin, Kathrin and Albedo.
At 14:00 first target images were taken by the AUPE WACs for testing science assessment.
In the meantime some Hazard maps evolved in the van. But VO still didn’t work.
Bernie, Alberto, Gerhard and Laurence tried all afternoon to fix the visual odometry bug – without success.
At about 16:00, the team decided to pack everything except Idris & network. Bernie & Alberto together with the AU team were left all alone on site to go on testing and bug-fixing (again without success). For this purpose, they changed Idris’ location to nearby the van which immediately attracted lots of tourists.
In the evening the plenary meeting took place from 17:00 to 19:30. Priorities, technical stuff and logistics as well as next day’s schedule were discussed. During the meeting, Frank, Laetitia (DLR) and Derek Long (KCL) joined the team.
After the meeting, work hadn’t finished for several team members. In the backyard people from AU, JR, GMV and SciSys tried to capture a test-dataset to be used for bug-fixing overnight / in the next morning (when not having access to Idris).
Sunset made people stepping inside and having dinner at 20:30. Due to VO still not working, it was going to be a long night…
By that morning two UCL Hypercams (HC, one visible and one near infrared) had been mounted on the PTU.
Steve, Mark & Fred went to site at 8:15 (about 30min) for setting up. Official beginning on site was said to be at 9:00.
The PTU was damaged by accidential weight, but it was solved by Laurence adding some offset in the morning (until 10:30).
Alberto (GMV) arrived on site at about 11:00.
At the same time a guy from the park authorities visited us and checked our permit.
Kathrin and Ernst took another walk along the trajectory to check the positions of the landmarks. Again some of them had been moved (by wind or Idris). One idea for improvement (for upcoming field trials) is to fix them with nails. Ernst took a look at potentially relevant regions of interest, but unfortunately no natural Martian like stones or sediment layers appeared along the traverse. So we hope for Derek’s targets.
In parallel, Bernie started running the vision SW (DEM processing and VO), where some problems appeared in saving TIFs. Alberto started testing mechanical odometry and did odometry calibration together with Fred (i.e. 5m driven = 4.98 measured by tape on flat ground, 20m driven = 19.939m measured by tape climbing).
At about 14:00-14:30 we had lunch in the van.
Afterwards Mark mounted the dGPS on Idris and together with Alberto they went on testing the mechanical odometry on site.
Laurence had a first look at the aerobot at about 15:00. Helium should be available by midday of 12th, but unfortunately there would be a delay of two more days.
Michal and Honza from CTU arrived at about 15:00 on site after having checked the Satellite Broadband connection to be provided by Electrosat on 16th and 17th. After a fast lunch, they visited the site for the first time and started to capture test images for computing the overall DEM.
Gerhard & Conny left the test site at about 15:00 to extend the aerobot permit by several days.
Kathrin & Ernst repositioned the shifted landmarks (1, 20, 63, 21, 16) by means of the existing GPS coos and belonging distances between several landmarks.
Bernie still tried to fix the vision SW problem, but due to nothing to do, people started packing up at about 16:00. At 00:09 the positive mail message was sent: “tifs can be stored”!
At about 20:00 Andy, Matt and Mark from SciSys arrived at the Parador and joined the rest for dinner.
In the morning, Conny from ActiveConnect (our local supporter) joined the PRoViScout team and discussed some administrative issues with Gerhard. Afterwards Conny and Kathrin prepared the meeting room at the Parador and updated the booking list.
After breakfast, Steve and Mark took a first look at the site (10:30).
Having a walk along the traverse, the JR team verified the landmarks still lying on the positions left the day before. Target nr. 6 was the only one to be “blown away” and had to be put back to its original position (known by image documentation + dGPS data).
Back at the van, Idris was “landing on Teide” (11:00).
Steve was mounting the PTU (Pan Tilt Unit) with HRC (High Resolution Camera) and AUPE (Aberystwyth University PanCam Emulator).
After about 2.5 hours of mounting and installation, everything was set up on site and the two Wifi networks (Idris - Fred, all - all) worked (13:30).
At about 13:30 Idris moved for the first time on the test-site (remotely) – and did very well. Soon first tourists appeared – simply having followed Idris’ tracks on the ground. Soon the decision was made to steadily remove Idris’ tracks leading to the site!
Communication on site was supported by a set of radios brought by AU.
At about 15:00, Conny joined us and we had lunch.
Afterwards Gerhard and Steve tried to calibrate AUPE. The results were not as good as expected, so they repeated the procedure in the evening on better conditions at the Parador.
Finally, at about 17:30, AUPE captured first images and sent them via network to the “control station” in the van. All in all the first day on site was very successful and the team left very pleased. The last thing to do was to clear the tracks…
From 19:30 to 20:30 a short meeting took place at the Parador’s meeting room. At that time Ernst from DLR as well as Andrew from UCL had already arrived. Several logistic and technical issues were discussed.
In the morning, Gerhard and Kathrin (JR) placed numbered white “landmarks” for geo-reference along the rover’s traverse. The coordinates were measured by dGPS. Gerhard took 50+50 backup images for 3D reconstruction and sent them to CTU for being processed (to be used as backup for the initial DEM).
In the late afternoon, the AU rover team arrives on Tenerife and clears customs (Santa Cruz). It was the first night for Steve and Mark (AU) to sleep in a comfortable bed since they left Aberystwyth.
The JR team left Austria from airport Graz at 8:25 (Austrian time) and arrived on Tenerife at 13:15 (local time). After arriving at the Parador at about 16:00, the first on-site visit took place.