Mars Descent Imager (MARDI) Update

Phoenix and MSL MARDI News

12 November 2007

This release addresses two recent developments concerning descent imaging on upcoming U.S. Mars landers. There is a good news/bad news component to the story.

First the bad news:

Phoenix MARDI will not take any images

Very late in the Phoenix lander ground testing program, a problem was discovered with a computer card in the Phoenix lander to which the descent imager is attached and which is used by the spacecraft computer to read the image data out of the camera. The Payload & Attitude Control Interface (PACI) card also provides the spacecraft computer access to the Inertial Measurement Unit (IMU) whose data are critical to controlling the descent and landing. Under anticipated and normal operations, the problem had benign consequences, but under specific conditions readout of the MARDI data could have interfered with readout of the IMU data during Entry, Descent, and Landing (EDL). This interference itself was only likely to create problems under additionally restrictive conditions, but could in fact lead the spacecraft computer to incorrectly respond to IMU input, causing Phoenix to crash. Since the MARDI contains a small (1 MByte) internal buffer, it was suggested that a single image be acquired and held within the camera until after the vehicle has landed, when IMU data are no longer critical. While not meeting all of the objectives of the MARDI investigation, a single image would have been extremely useful from both a scientific and mission planning perspective. Phoenix was launched with the intent to study the issues associated with acquiring a single descent image.

After launch, MSSS, Lockheed Martin, and JPL began examining how a single image could be acquired and returned. In order for the one image to be worthwhile, several things needed to happen: since the camera would not have taken any images for some time, we needed to flush away charge that had accumulated on the detector, then the actual science exposure needed to be taken, and its read out inhibited, and then the image needed to be read from the camera's internal memory. Additionally, one or more image taken on the ground after landing would provide calibration information. This was a non-standard way of operating the camera. Although nearly all of the changes that needed to be made were internal to the MARDI (and if they didn't work, we'd just not get an image), small changes in the EDL sequence which controls the entire landing process were also required. Project engineers argued that of the two options (changing the sequence so that the MARDI was not powered on or changing the timing of events within the EDL sequence to take one image 19 seconds after the landing engines began firing and then reading it out after landing), not powering MARDI on was the lowest risk. Although the risk to landing of either change was deemed very low, changing the timing within the EDL sequence would have necessitated additional testing, and arguments were raised that the human and time resources for this testing were in short supply. Reluctantly, the Phoenix Principal Investigator (Peter Smith) agreed, and the decision was made to leave MARDI powered off and not to take any descent images.

For those who were also expecting to hear audio from the MARDI's microphone, this decision also eliminates that possibility.

Now for some good news:

Mars Science Laboratory (MSL) MARDI re-instated to Flight Status

As part of an effort to address cost over-runs within the MSL Project, NASA elected to remove the MARDI from the MSL spacecraft. However, the cost savings associated with this decision were very small. Since the MARDI shares the exact same electronics as three other cameras on the MSL spacecraft (two Mastcams and one Mars Hand Lens Imager, MAHLI), and the electronic parts and boards had already been procured, along with the glass elements for the lens, the only remaining costs to complete and test/calibrate the flight camera was a few $10,000's. A bigger but still very small cost was that needed to integrate and test the camera with the spacecraft. These latter costs would include the labor hours of engineers and technicians to test the camera in a stand-alone setting, then to physically attach the camera to the MSL rover and connect it electrically, and then to test the camera functionally. Also included in this cost would be the software necessary to command the instrument on and off, and to read out its data. This is estimated to be several people-months of work effort, split between probably eight to ten people, or more than $100K.

As the small magnitude of these costs became more apparent, Mike Malin (MARDI Principal Investigator) worked closely with the Mars Exploration Program (MEP) at JPL and Mars Program at NASA HQ to identify ways of offsetting these costs. Malin offered to pay the costs of completing the camera at MSSS, but the JPL costs of integrating the camera remained. When the Phoenix Project made the decision not to take any MARDI images, the MEP inquired of the Phoenix PI (Peter Smith) if the funds allocated to Phoenix MARDI could be reallocated to MSL. With the provision that some resources be retained to enable Malin to participate as a science team member in Phoenix operations, Smith agreed to reallocate funding from Phoenix to MSL to cover the JPL costs.

With the primary concern raised about MARDI addressed, the NASA Associate Administrator for the Science Mission Directorate agreed to return MSL MARDI to flight status. MARDI will be delivered in May 2008, and will take 1600x1200 pixel RGB color images at a rate of 5 frames per second during about 2 minutes of descent and landing in September 2010.

Why is Descent Imaging Important?

Within the context of its recent challenges, there has been a natural tendency to ask whether or not descent imaging is worth the effort. Indeed, some have argued that in light of the extremely good, high spatial resolution images being taken by the High Resolution Imaging Science Experiment (HiRISE) on the Mars Reconnaissance Orbiter (MRO), perhaps descent imaging isn't needed any more. The primary objectives of descent imaging are to identify where the lander has landed, and to bridge the gap between orbiter science views and those acquired by the lander. Since HiRISE images taken after landing can show where the lander is (e.g., VL1, VL2, MER-A, MER-B) and some small features seen in the HiRISE images match those seen in lander views, it is possible to argue that the need for descent imaging can indeed be met without flying dedicated cameras.

However, we do not know how soon after landing HiRISE will in fact be able to image a lander, and a number of time-critical mission functions still would greatly benefit from having an overhead view of the landing site immediately after landing. Additionally, as seen in the accompanying images, there is considerable science and engineering information in more detailed images than can be discerned in HiRISE resolution and quality data. MSL MARDI will take hundreds of images that will start at HiRISE resolution (several 10's cm/pixel scale), and then increase to a final scale of a few mm/pixel.


Synthetic Comparison of HiRISE and MARDI Image of Representative Location within Phoenix Landing Site

(Antarctic Aerial Photograph of Polygonal Ground superimposed on HiRISE Image of Martian Northern Plains)


Animated Comparison of Simulation of HiRISE and MARDI Image Resolution

(Right image is enlargement from middle of Left image)

However, perhaps the most important function of descent imaging is terminal hazard avoidance (i.e., actively steering the lander to miss hitting rocks or steep slopes). Neither Phoenix nor MSL are employing hazard avoidance, and with respect to surface hazards, their landings are statistical "crap-shoots." Considerable effort is expended by the MEP, the Mars science community, and each Project to find places that reduce the risk of hitting something geological, but these efforts generally compromise the science return of the mission and remain statistical. With a demonstrably effective hazard avoidance system, a successful landing would be much more likely. But the key word is "demonstrably." Although hazard avoidance has been proposed for previous missions (including Phoenix and MSL), engineering review panels have been very conservative about allowing any flight use of such capabilities. Somehow they convince themselves that the risk of trying to actively steer the lander using input from hazard sensors introduces greater risk than not trying to avoid the hazards.

What does this have to do with MSL MARDI? Those of us who believe that active hazard avoidance is necessary to be able to land at the most important science sites feel we will need a real-world data set, acquired under actual flight conditions, to test the software central to hazard avoidance systems sufficiently to address the objections raised by past engineering review panels. MSL MARDI will take hundreds of images in a near-video-like sequence that can be used in such tests. Mars sample return and autonomously landed components of the human exploration program on both the Moon and Mars will likely need such capabilities.