Emily Lakdawalla

by The Design

  Okon A (2010) Mars Science Laboratory Drill. Paper presented to the 40th Aerospace

  Mechanisms Symposium, 12–14 May 2000, NASA Kennedy Space Center

  Sunshine D (2010) Mars Science Laboratory CHIMRA: A device for processing powdered

  Martian samples. Paper presented to the 40th Aerospace Mechanisms Symposium,

  12–14 May 2010, NASA Kennedy Space Center

  6

  The Mast, Engineering Cameras, Navigation,

  and Hazard Avoidance

  6.1 INTRODUCTION

  The Curiosity mission navigates Mars using a combination of human and artificial intel-

  ligence. Both methods rely upon a suite of engineering cameras for situational awareness.

  The twelve engineering cameras are in six pairs: two redundant pairs each of Navcams,

  front Hazcams, and rear Hazcams. A remote sensing mast lifts the four Navcams nearly

  two meters above the Martian surface, while the eight Hazcams are mounted at belly

  height, four facing forward and four to the rear. The Hazcams and Navcams are flight

  spares or build-to-print copies of the engineering cameras of the same names on the Mars

  Exploration Rovers; this not only saved money in hardware, but made it significantly eas-

  ier to use a modified version of the same rover driving software for Curiosity as for Spirit and Opportunity. The mast also carries the Mastcams and parts of the ChemCam and

  REMS instruments. Both Navcams and Hazcams are routinely used to gather data for

  environmental science purposes.

  6.2 REMOTE SENSING MAST

  Curiosity’s vantage point is a bit higher than most humans’. From the Navcams’ position

  at 1.9 meters above the Martian surface, Curiosity can see quite far: if the landing site

  were perfectly flat, the horizon would be 3.6 kilometers away. Of course, Curiosity sits

  inside a crater, and topography rises above the horizon as far as the rover can see. The

  nearest foothills of Gale crater’s central mound were about 5 kilometers from Curiosity

  on landing day. The nearest point on Gale’s rim was 20 kilometers to the north; to the east and west, the visible rim is more like 40 kilometers away. All of this topography is usually visible in Curiosity images of the horizon, although the crater rim and sometimes

  even the central mountain disappear and reappear over time as the amount of dust in the

  air waxes and wanes.

  © Springer International Publishing AG, part of Springer Nature 2018

  221

  E. Lakdawalla, The Design and Engineering of Curiosity, Springer Praxis Books,

  https://doi.org/10.1007/978-3-319-68146-7_6

  222 The Mast, Engineering Cameras, Navigation, and Hazard Avoidance

  The Remote Sensing Mast (RSM) has three motors, of which one, the mast deploy

  actuator, was used only once to lift the mast permanently to its vertical position.1 The

  mast’s azimuth and elevation actuators are mechanically capable of panning 362° hori-

  zontally and tilting 182° vertically in order to point cameras at every possible target

  within Curiosity’s view. The elevation mandrel and azimuth twist cap allow the cabling

  to flex as the mast tilts and rotates. Software prevents the mast from rotating to hard

  stops, limiting it to panning 360°. Software also limits how far down it can tilt, prevent-

  ing it from pulling on the cable bundle that runs up the mast to the mast head, reducing

  its tilt limit by 4°. Thus its tilt can be commanded from –87 to +91°. Two booms contain-

  ing REMS instrument components are mounted to the mast below the actuators, so their

  positions are fixed (Figure 6.1).

  Figure 6.1. Parts of the remote sensing mast as seen in photos taken during assembly at JPL

  in 2011. The mast is about one meter tall. In this photo, the two REMS booms are covered for their protection; the covers were removed before flight. NR-A, NL-A, NR-B, and NL-B refer to the right and left Navcams connected to the A-side and B-side computers. NASA/JPL- Caltech/

  Emily Lakdawalla.

  1 The mast and engineering cameras are described in Maki et al. (2012)

  6.2 Remote Sensing Mast 223

  The mast has to be able to point incredibly precisely in order for ChemCam to zap

  targets selected within Navcam images. The mast’s absolute pointing accuracy is 0.25°

  (4.6 milliradians), or about 6 Navcam pixels; but its pointing is repeatable to less than a single Navcam pixel. This pointing precision has enabled Curiosity to perform sharp-shooting feats like profiling down the side of drill holes (Figure 6.2).

  Figure 6.2. ChemCam laser shots are spaced using minute motions of the remote sensing mast. Here, ChemCam laser shots are spaced 1.4 mm apart horizontally in rasters marching down the wall of the Okoruso drill hole. Drill holes are 1.6 centimeters in diameter, and this one was 2.6 meters away from the mast head when the shots were fired. Corresponding mast shot-to-shot angular motion was 0.03° (0.5 milliradians). MAHLI image

  1338MH0005880000501506R00, taken at night with LEDs on. Contrast in the image has

  been increased to emphasize the laser shot points. NASA/JPL-Caltech/MSSS/Emily

  Lakdawalla.

  224 The Mast, Engineering Cameras, Navigation, and Hazard Avoidance

  6.3 ENGINEERING CAMERAS: NAVCAMS AND HAZCAMS

  The Navcams are located approximately 1.9 meters from the ground when the rover is

  level, but their precise elevation depends upon the tilt of the mast head. They are spaced a very wide 42.4 centimeters apart, which gives them depth perception out to a distance of

  100 meters. That helps engineers plan long blind or visual odometry drives, the most time-

  efficient driving modes, as long as the Navcams have a good view of the path ahead.

  Figure 6.1 shows the positions of all the Navcams. The switch from A-side to B-side cameras after sol 200 moved the rover’s Navcam vantage point downward by 4.8 centimeters.

  The fish-eye Hazcams provide Curiosity with situational awareness of the terrain both

  forward and aft of the rover and in between the wheels, particularly in areas not visible to the Navcams. The Hazcams are hard-mounted to the rover so have fixed fields of view.

  Figure 6.3 shows the locations of all the Hazcams. They are boresighted 45° below the horizon, with 120° field of view vertically, and 180° corner-to-corner. Because of the wide view, raw Hazcam images are very distorted. The front Hazcams are mounted near the

  middle of the front of the rover, with the A-side and B-side cameras interleaved, each off-

  set from the next by 8.2 centimeters, giving a stereo separation of 16.4 centimeters for

  each pair. The front Hazcams provide detailed stereo maps of the area within reach of the

  robotic arm.

  When Curiosity switched to the B-side computer, the front Hazcam view of the world

  shifted to the rover’s left by 8.2 centimeters, resulting in a view that is more obscured

  (primarily by the shoulder elevation actuator) than the previous view was (Figure 6.4). The rear Hazcams are mounted on either side of the RTG in two pairs spaced 10 centimeters

  apart, with the A-side rear Hazcams on the rover’s left side and the B-side rear Hazcams

  to the rover’s right, separated by 1 meter. So when Curiosity shifted to the B-side com-

  puter, the rear view seemed to shift left by 1 meter. The new rear view is not substantially different in quality from the old view.

  The Navcams and Hazcams have identical detectors, 1024 pixels square. It takes 5.4

  seconds to read out a single full-frame image. Rover planners can improve that speed by

  binning the images or by reading out partial, “windowed” (cropped) images. The cameras

  are sensitive to light i
n the 600 to 800 nanometer range – slightly longer-wavelength than

  human color vision, and similar to the red filter on the Mastcams (Figure 6.5). Only two of the six cameras can be powered simultaneously. So stereo pairs are usually taken at the

  same time, but front and rear Hazcam pairs have to be taken sequentially.

  Navcam images of the path ahead and arm workspace are decisional data, required to

  plan later sols (see section 3.3). It can be tricky to squeeze all the necessary data into the first available downlink, especially if a drive happens late in the sol after the Mars

  Reconnaissance Orbiter communications pass so only an Odyssey pass is available.

  Compressing the images reduces file sizes, which allows more images to be returned to

  Earth. But lossy compression reduces data quality, potentially affecting the quality of the range information that rover drivers use to plan driving and arm positioning. So images

  that are used for generating range maps are compressed very little, while other images

  taken only for documentation purposes (for instance, to verify the placement of the robotic arm) are compressed much more.

  6.3 Engineering Cameras: Navcams and Hazcams 225

  Figure 6.3. Locations of the Hazcams. Top image taken during JPL mobility testing on 3 June 2011. NASA/JPL-Caltech image release PIA14254. Bottom image taken at arrival of the

  rover at Kennedy Space Center. Credit: NASA/Frankie Martin, release KSC-2011-5909.

  226 The Mast, Engineering Cameras, Navigation, and Hazard Avoidance

  Figure 6.4. Shifting Hazcam points of view between sol 166, on the A-side cameras, and sol 233, on the B-side cameras. The rover did not change position in the time between these two sets of images. The images have been reprojected to correct for the fish-eye distortion of the Hazcams. NASA/JPL-Caltech.

  6.4 Using the Engineering Cameras 227

  Figure 6.5. Spectral responsivity of Navcams and Hazcams. From Maki et al. ( 2012 ).

  6.4 USING THE ENGINEERING CAMERAS

  6.4.1 Navcam panoramas

  With the Navcams’ 45° field of view, it technically takes only eight pairs of Navcam

  images to cover the complete 360° in stereo around the rover. But to generate good ste-

  reo range information for planning, it’s necessary to have substantial overlap between

  adjacent image tiles, so most 360° Navcam panoramas contain 12 image footprints

  (where a footprint includes one each of left- and right-eye images) in any tier. Curiosity

  requires so much overlap because the wide spacing of Curiosity’s Navcams translates

  into a 21-centimeter offset from their pan axis. Rotating the mast shifts the camera posi-

  tion, making edges of adjacent Curiosity Navcam frames match poorly, particularly

  close to the rover.

  Occasionally, particularly at drill sites, the rover takes a complete lower tier Navcam

  panorama to image the deck. The top of Mount Sharp is usually cut off in standard Navcam

  panoramas, but occasionally the team commands an upper tier to fill Mount Sharp in.

  When the rover is traveling in valleys among ridges or buttes, the team may command

  partial or complete upper tiers (often just half of the Navcam field of view) to capture

  topography above the horizon.

  228 The Mast, Engineering Cameras, Navigation, and Hazard Avoidance

  6.4.2 Drive imaging

  Most drives end with two high-priority Navcam panoramas, crucial for planning the next

  sol’s activities. One is a 5-by-1 array of stereo pairs that covers the likely future path of the rover, up to and just above the horizon: the “drive-direction panorama.” Another is a 5-by-1

  array pointed off the front right corner and right side of the rover, one tier down from the drive-direction panorama: the “ChemCam targetable region.” When data volume permits,

  the rover acquires a complete 12-frame, 360° panorama after a drive by adding in left and

  right “wings” of two frames each and then the rear view, comprising the last three frames.

  Because much of the rear Navcam view is occluded by the RTG and UHF antenna, the rear

  view isn’t very useful for drive planning, so the rear-view images have much lower down-

  link priority than all the rest. As a result, they are often not returned to Earth until many hours after the rest of the panorama. On sols when resources are limited, the rear-view

  portion of the panorama may be deferred until the next sol, or not taken at all.

  When the rover uses Navcams and Hazcams for visual odometry or autonomous navi-

  gation, it takes them in a 4-by-4 summation mode, producing images only 256 pixels

  square. Visual odometry frames look like the view out the window of a moving vehicle,

  with rocks and other features slowly tracking across the field of view. Autonomous naviga-

  tion adds Hazcam frames to the mix, interleaved with the Navcam images. If mid-drive

  Hazcam images are full resolution (1024 pixels rather than 256 pixels square), that’s usu-

  ally a sign of mid-drive use of the DAN instrument in its active mode, rather than autono-

  mous navigation (see section 8.3).

  6.4.3 Slip checks

  Even if engineers don’t plan to move the rover, they usually command Hazcam imaging as

  the first activity of the day, to make sure that thermal contraction during the overnight chill hasn’t caused any shift in the rover’s position. Slip-check images are also useful after arm activities, because the arm’s substantial weight can cause the rover’s position to shift

  slightly.

  6.4.4 Environmental observations

  The meteorology science theme group frequently uses Navcam movies for routine obser-

  vations of atmospheric dynamics. There are two main types: zenith movies and Mount

  Sharp movies (technically called supra-horizon movies).2 Both require only rover-relative pointing so can be performed on restricted sols. They are simple to command and produce

  low volumes of data, so can be captured during periods when the rover needs to be rela-

  tively inactive (e.g. over conjunction and lengthy Earth holidays).

  The team takes zenith movies to search for high-altitude clouds. To capture zenith mov-

  ies, the Navcam points at an elevation of 85°, almost directly overhead. A Navcam shoots

  8 images at intervals of about 13 seconds, observing for a total of 91 seconds. The images

  are downsampled by a factor of two, producing 512-pixel-square images. To analyze the

  2 Kloos et al. (2016)

  6.4 Using the Engineering Cameras 229

  images, the atmospheric science team averages the 8 images together and then subtracts

  the average frame from all 8 original frames to search for faint ghosts of clouds in each

  image. If the Sun were in the field of view, it would overwhelm the Navcams’ ability to see clouds. Therefore, the rover never takes zenith movies within 3 hours of local noon; and

  takes most in the late afternoon. To further avoid the Sun, the Navcam points north to take photos during the winter (Ls 0–180) and south during the summer (Ls 180–360). On average, the mission acquires these observations about once every 6 sols.

  Mount Sharp movies are to search for orographic clouds over Mount Sharp. They can

  also reveal lower-altitude clouds because they look at a lower angle through the atmo-

  sphere than zenith movies. To take them, a Navcam points southeast, at 135°, at an eleva-

  tion of 38.5°. To avoid the Sun, these movies have to be taken after 10:00 a.m. local solar time. Initially, they were taken the same way as the zenith movies (eight frames, 512 pixels square, at intervals of 13 seconds), but after sol 594 the sequence and pointing was changed to cover more of the mountain and ground i
n a swath 1024 pixels tall by 512 pixels wide.

  To keep the data volume the same, they reduced the movies to only 4 frames captured at

  intervals of 13 seconds.

  There are also dust devil movies, in which the rover gazes to the north to search for the

  motion of dust devils across the plains. 3 The northward direction was chosen because it offered Curiosity the longest-distance view in which dust devils might be visible. Dust

  devils were observed in only two of 250 dust devil movie observations. As it turned out,

  dust devils were happening, but the Navcams were pointed in the wrong direction to see

  them. On sol 1520, a dust devil was fortuitously spotted in a Mastcam multispectral obser-

  vation aimed at Mount Sharp. Since then, the environmental science theme group has

  aimed dust devil movies toward Mount Sharp at the south and observed lots of them

  marching across the lower slopes of the mountain. 4

  A particularly pretty type of Navcam observation is Navcam sunset movies, to deter-

  mine scattering properties of the atmosphere.

  6.4.5 Anomalies

  The switch from A-side to B-side cameras after the sol 200 anomaly should have been a

  relatively minor event. Unfortunately, the rover planners found after the switch that the

  terrain meshes derived from A-side and B-side cameras did not match. Engineering cam-

  era team lead Justin Maki figured out that the camera bar to which the Navcams are

  mounted warped with temperature change.5 The engineers had to develop a temperature-dependent camera model and upload it to the rover before they could use autonomous

  navigation capability.

  Images from the rear Hazcams often appear significantly noisier than those from the

  front Hazcams. The rear Hazcams run much warmer than the front ones due to their prox-

  imity to the hot MMRTG radiator fins. The high temperature increases the cameras’ dark

  current, amplifying the brightness of hot pixels.

  3 Moores et al. (2014)

  4 Lemmon et al. (2017)

  5 Justin Maki, personal communication, review dated September 22, 2017