ESTEC-CNES ROVER REMOTE EXPERIMENT

Transcription

1 ESTEC-CNES ROVER REMOTE EXPERIMENT Luc Joudrier (1), Angel Munoz Garcia (1), Xavier Rave et al (2) (1) ESA/ESTEC/TEC-MMA (Netherlands), (2) Robotic Group CNES Toulouse (France), ABSTRACT In June 2010, a remote testing has been performed between ESTEC and CNES with the purpose to exercise ExoMars rover-like operations in a realistic environment with limited budget. The benefits are multiple: operational set-up equivalent to real operational conditions, no travelling cost, no impact from weather (need only of reschedule), allows participation to more people (scientists, engineers etc ). Thanks to the interface and rover prepared by CNES, the first experiment was a success. This paper discuss the benefits of such testing and present the results of the first test that allowed driving about 70m on the CNES Mars Yard called SEROM. Plans for future tests is discussed with the possible enhancement of the tools used. 1. INTRODUCTION Rover field testing is a systematic activity for NASA, as part of a mission preparation or as part of R&D and operational validation. It is also used for general confirmation of the understanding of rover operations associated with particular scientific instruments. ESA does not have such systematic approach. A first step toward rover field testing has been performed using CNES-provided infrastructures that hopefully will lead to more systematic rover testing at ESA. In June 2010, a remote experiment as been conducted between ESTEC and CNES. It consisted in commanding IARES rover on the SEROM Mars Yard from ESTEC in conditions similar to a real martian mission. It was focused on commanding the rover to travel toward a specific goal with the objective to experiment and assess the necessary tools and operational sequences. and conducting tactical planning activities to upload new rover activity commands in only a few hours. For the traverses, coordinates of the targets (intermediate and final) are defined by ground, together with possible stay-out zones and parameters of the autonomous navigation function on-board such as maximum allowable slope to be traversed or maximum step height to be negotiated [1]. The basic assumptions are that the rover sends a panorama (e.g. using PanCam instrument) to ground together with its estimation of its localisation relative to the previous travelling command. Based on these data, the ground operators shall estimate the true absolute position of the rover on a map, register the PanCam images (the absolute attitude of the rover including heading is estimated with error by the rover). Another assumption is that images and Digital Elevation Maps (DEM) of the terrain are available on ground (e.g. HiRISE images and DEMs [6]). This is currently a reasonable assumption as the landing ellipse would be precisely characterised during the landing site selection process [5] with available remote sensing capabilities SEROM Mars Yard The SEROM Mars yard is an artificial outdoor terrain with rocks and slopes of about 80m long by 50m large. The soil and rocks are made of lava rock debris (puselan) and can be configured where needed. Prior to the campaign, CNES performed a precise mapping of the terrain enhancing the resolution from about 40 to 30cm/pixel view publicly available to about 5cm/pixel. This newest view was made available in an improvement of the graphical interface of the rover commanding tool (ANW see below). 2. EXPERIMENT SET-UP ExoMars-like operations [1,2] are meant to be conducted on a daily basis, receiving telemetry from the previous sol (a sol is a martian day), evaluating them Figure 1 Old and new view of the SEROM terrain.

2 2.2. IARES rover The IARES rover is the development platform of CNES since many years (see figure 3). It is equipped with a pair of navigation stereo cameras on a Pan&Tilt mechanism fixed on a mast about 2m above the ground. The rover has 6 driving wheels and a complex passive and active locomotion subsystem. In this experiment, only simple locomotion capabilities have been used. IARES is equipped with a GPS allowing precise absolute localisation. Its on-board computer is running the autonomous navigation functions until the defined target has been reached [3, 1]: stereo image processing, navigation maps creation, path planning, and trajectory execution Autonomous Navigation Workshop (ANW) The Autonomous Navigation Workshop is the command interface for the rover. It allows setting the target s coordinates, parameters of the autonomous navigation algorithms and start/stop the rover. The coordinates can be entered in different ways: manually giving the X&Y absolute coordinates or graphically, by clicking on the map (figure 2). As said previously, CNES has improved the resolution of the map using an up-to-date image of the SEROM. Terrain features can be seen with quite some details. An additional valuable feature of the interface is the position update of the rover as well as the trajectory display. We will discuss their adequacy and use in the operations process later in the paper. Figure 3. IARES on the SEROM from Webcam 2.5. Experiment architecture The architecture of the experiment is fairly simple: CNES personnel have placed the rover on the terrain and switched it ON. During the whole experiment, CNES personnel monitored the engineering parameters of IARES to ensure health. They typically interrupted the test after few hours to change the batteries of the rover. They stayed on the phone to be aware when ESTEC intended to command the rover. On ESTEC side, one operator was in charge of commanding the rover using ANW remotely connected. Another operator was controlling the webcam, in charge of taking pictures regularly to document the test. He was also helping in some image post processing. Finally, another operator was supervising the experiment, specifically defining the target of interest. We had the intention to invite ESTEC scientists to be part of this experiment. We wanted to have them in a separate room where they would only have the end products (panorama and rover position estimation on the map) like in a real mission. Due to availability reasons under short notice, this was not achieved but it should be one of the objectives for next experiment. Figure 2. ANW GUI 2.4. External Webcam & Phone A webcam mounted on a Pan&Tilt mechanism is also available. It can be remotely commanded including its zoom. It provides a valuable global view of the scene although such view would not be available in a real case (see figure 3) A telephonic line was also available during the conduction of the experiment between ESTEC and CNES operator monitoring the rover. 3. PREPARATION 3.1. Target coordinates determination One of the objectives of the test was to define the coordinates of a target and understand from a ground operator perspective the situational awareness difficulties after traversing some distance in a poorly known area (like for a real mission). Although the ANW graphical interface was enhanced by CNES with a precise map of the terrain, we wanted to use simpler means to assess the necessity of advanced tools development. Coordinates relative to the rover can be computed from

3 pure X&Y coordinates or from a heading and a distance. Since the panorama of the terrain is meant to be used by scientists in conjunction to other orbital remote sensing data, we have tried to use only those data. We have used a free computer program to stitch the initial panorama images together. Then we have added vertical lines to help the operator to understand the heading information of the images (see figure 4). The heading knowledge is essential to link panorama features to the features that can be also seen on the map. Figure 4: Stitched panorama with heading cues. Such need obviously calls for an immersive environment to enhance this situational awareness issue. We have attempted to use a DEM of SEROM kindly provided by Joanneum Research [7] into the virtual environment of 3DROV available at ESTEC [4]. However, when fitting the image of the SEROM on this DEM, we could see some inconsistencies that couldn t be resolved in a timely manner. Anyway, since SEROM had changed since the stereo images acquisition were made (june 2008), the DEM was of minor use for our experiment. However, the need of a full DEM acquisition campaign of the SEROM was identified at a precision equivalent to HiRISE DEMs [6]. Figure 5: SEROM course DEM Definition of the target coordinates was intended by identifying the interesting feature on the panorama (grey level instead of multi spectral images intended by PanCam). Then computation of the heading of the feature in the image (easily deduced from the distance to the zero degree vertical cues added on the panorama). The heading direction was then reported on the map to define the distance to the target. In case we would have had PanCam stereo images, the distance could have been estimated using stereo matching or DEM construction. Unfortunately, our stereo image of the navcams could not support this function with only 10cm stereo baseline (7cm distance). In case long traverses would be needed (initial intention of ExoMars is 100m per sol), the use of stereo images would not be adequate, or other methods should be developed like using larger stereo baseline by moving the rover aside. Another option would be to use the ANW GUI image with 5cm/pixel resolution. Identifying the target feature or a close feature recognisable both on the map and on the panorama was the easiest way but was not actually representative to a real case because ANW map is too accurate (an ANW enhancement could be to allow easy down sampling of the map). We have finally estimated the distance to the target by selecting recognisable large features on the map and on the panorama, near the intended target. This was prone for larger errors. Using the GUI of ANW, we could estimate that we made less than 2 meters error for the first target about 30m away. This was considered acceptable. In addition, we wanted to understand what would be the effect of such approximations to the operations Operations procedures & rehearsal One of the key for the success was clearly the availability of a very detailed procedure rehearsed a number of time prior the test. Even if the ground process was not very complicated, this was necessary to be efficient. This experiment was meant to demonstrate the feasibility of such tests. Therefore, no effort had been spent in ground process automation. The main steps were: 1. connect to the rover 2. take a panorama images, 3. stitch the images and add heading features 4. select a scientific target 5. estimate the target coordinates 6. compute the coordinates in the proper format and command the rover In parallel, an operator was taking the webcam images. The file naming convention, saving directories had been agreed prior to the test. The procedure had been rehearsed several times off-line prior to the actual experiment during verification of the connection with the rover. 4. REMOTE TEST In June 2010, the test was conducted with the set-up described above. CNES had provided the absolute heading of the rover as well as the initial position of the rover on the SEROM (figure 8) were we started to take the initial panorama.

4 4.1. First target Figure 10: first target Figure 6: Initial panorama Figure 7: First target The first target was reached, thanks to the autonomous navigation functionalities on-board IARES. Obviously, the new panorama is showing a new place that was poorly known. It is difficult to recognize features from the initial panorama. Here also, a virtual environment, possibly immersive would be justified. Again, Coregistration of the panorama with respect to heading showed to be important. The heading cues that we have been adding are very helpful but are relative to the rover heading. Additional absolute heading cues would have been necessary to identify features in various panoramas Second target From the panorama at the first target, a new target was chosen. Circled of red is the largest feature near the scientific target of interest circled of green in figure 11. This later was chosen because of the very bright white spot. The yellow cross was chosen to be the approach target and was estimated 25m away. Figure 8: ANW GUI with selected target location It is clear that for an operator, the link between the features on the panorama and the map is not obvious. Beyond few tens of meters from the rover, the resolution of the image decreases a lot. The initial target distance was estimated 28m away from the rover. It was deliberately chosen to go on the right side of the rockfree corridor which would not be stressing the autonomous navigation enough. Figure 11: zoom on the second target Figure 9: traverse to the first target Figure 12: traverse to the second target

5 Figure 13: second target At this point, operators at ESTEC were not sure anymore of which feature was representing the second target. Uncertainties between the target identified on the panorama, on the map and the believed position and heading of the rover were leading to guess work on which feature was corresponding to the intended scientific target (green circled) and recognisable near feature (red circled) in figure X. It was decided to come closer to the finally identified features below. White scientific target revealed to be some vegetation. Figure 14: Identified scientific and near feature targets Final approach It was decided to move 2 meters closer to the scientific target as a rehearsal of a target approach. Figure 15: final approach An image of the scientific target was taken and a navigation map of the area surrounding the rover. The later is difficult to interpret and would need tools to coregister and define which the scientific target is. Figure 16: image of the scientific target and final navigation map 4.4. Some figures The experiment took about 3.5h hours to reach the 55m distance in three steps (28m, 25m and 2m). The rover took 20minutes to reach the first target. Ground took 35minutes to take the panorama and define the next target. The rover took 20minutes to reach the second target. There is clearly room for improvement though automation of the ground process. All the data have been stored. A post processing tool allowing view and assessment of the traverse (navigation maps) at each navigation step has been developed by CNES but was not used in the frame of this experiment. The navigation map was mainly needed to assess distance to the final target; otherwise, the autonomous navigation function was working well, bringing the rover where expected. Note that in this simplified experiment, the estimation of the resources to traverse was completely out of scope. 5. POSSIBLE FUTURE TEST This initial test demonstrated the feasibility at low cost of field test gaining experience in the operation of planetary rover mobility. Future tests should incorporate missing elements such as a PanCam breadboard, as the eyes of the scientists. Stereo should help estimating distances and colour should help in the target determination and discrimination. The tools should be improved using a unique virtual environment where co-registration of images and DEM would ease the situational awareness as well as target determination. The resolution of the data (orbital images and DEM) shall be in line with HiRISE data so that higher resolution cannot influence the process. The external webcam is valuable for test documentation. Although it was not really used during the ground process, next test should have such monitoring outside the control room. Scientists participating in the target determination process would also help to grasp requirements from the intended end-users.

6 Future tests should also address the maximum capabilities of the mobility system to assess if special tools and procedures should be developed. 6. CONCLUSION This experiment was performed with minimal means during an afternoon of june 2010, thanks to the framework put in place by CNES allowing remote control of the IARES rover. Initial experiment was postponed because of technical issues and weather, but it did not impact anyone as no travelling of people or material was involved. The success of the experiment on ESTEC side, was mainly due to a thorough preparation and rehearsal of the procedure. The experiment has put in evidence necessary improvements on the ground processing tools, including some level of automation. This experiment should call for more systematic testing of the tools developed in Europe for ground control of rover missions. Finally, the CNES infrastructures (SEROM, rover, ANW software, external webcam, and telephone) are adequate for a first set of testing where we can learn a lot with minimal cost implications. Then rover testing in Europe shall evolve toward real field testing to encounter a larger variety of terrains and operational situations. 7. REFERENCES 1. L. Joudrier et al. (2009). Challenges of the ExoMars Rover Control. infotech@aerospace 2. R. Trucco et al. (2008). ExoMars Rover Operation Control Centre Design Concept and Simulations, ASTRA R. Rastel et al. (2008). Autonomous Navigation : a development Roadmap for ExoMars, ASTRA L. Joudrier et al. (2010). Overview of 3DROV, A planetary exploration rover design and verification tool, MODPROD Workshop on Landing Sites for Exploration Missions 6. High Resolution Imaging Science Experiment HiRISE 7. PRoVisG