Spring 2004 - OSU College of Engineering
Cover Story: Mapping Mars
By Geneva Ringel
Since January, Ron Li, professor of civil and environmental engineering and geodetic science, has been shuttling between the Ohio State campus and the Jet Propulsion Laboratory in Pasadena, California. The reason for Li’s hectic travel schedule is his participation on the 2003 Mars Exploration Rover Mission science team.
“I’m dealing with three different time zones - Eastern, Pacific and Martian,” said Li.
Regardless of the time zone he’s currently in, Li has been busy over the course of the mission.
 |
A group of Ohio State researchers, led by Li, have been conducting tests of an advanced mapping technology on the surface of Mars. Li’s research team includes research staff Dr. Kaichang Di and graduate students Fengliang Xu, Jue Wang, Xutong Nin and Charles Serafy.
Li and his research team have developed a software system, called OSU MarsMapper, which analyzes photographs taken by the rovers to provide valuable information about the terrain of Mars and the location of the rovers. Over the course of the mission, the Ohio State technology was tested alongside the currently used rover localization system.
Not all of the mission was experimental. The Ohio State mapping technology was used to create a 3-D photorealistic model that helped determine the initial landing position of the rovers.
Remote-Controlled Exploration
The semi-autonomous robots, Spirit and Opportunity, are the latest evolution in technology for the remote exploration of planets.
For the 2003 Mars mission, rover operations were controlled by instructions sent only once a day from mission headquarters millions of miles away. The rovers were equipped to avoid some hazards, but precise navigation was required to carry out operations in an efficient manner. Engineers needed to know the position of the rover before applying instructions for the next day’s operations.
Unable to rely on GPS or human observation, mission engineers must estimate the location of the rovers based on information gathered by the rovers’ on-board sensors. The primary localization system for the rovers used information collected from wheel odometers, an accelerometer, a gyroscope, and a sun sensor to calculate rover motion and location. This method, however, is susceptible to a drift in position accuracy over long distances. In particular, rock bumping and wheel slippage can result in inflated odometer readings, especially in loose sandy soil, like that found on Mars.
Beginning in 1998, Li and his team began working on a technology, in conjunction with the Jet Propulsion Laboratory, which could more accurately determine the position of semi-autonomous robots like the rovers.
 |
The resulting Ohio State MarsMapper software utilizes a mapping technique called photogrammetry. With this technique, stereo photographs taken by robotic explorers on Mars are analyzed to provide rover position and trajectory as well as information on terrain features that can be used to produce high resolution topography maps.
Prior to the 2003 Mars Mission, the software was tested on rovers in the Mojave Desert in California. During these tests, GPS data was available to test the accuracy of the technology. The results showed that the Ohio State MarsMapper software significantly increased the accuracy of previously used rover localization technology. The Ohio State technology reduced the rate of error from 10 to only 0.1 percent.
Everyday the research group would analyze 20 to 100 pictures. The rover took traversal photographs as well as a series of panoramic photographs covering a 360 degree area each time it stopped. For each sequence of photographs, the researchers “stitched” the photos together to create an image network containing all the camera views. Common tie points, such as rocks, from photograph overlap were used to stitch together the model.
According to Li, the Ohio State MarsMapper software automatically selected terrain features for tie points based on improved interest point filtering and image matching. This innovative approach significantly speeds up an otherwise tedious task.
Complex algorithms written by Li and his group were then used to produce information on the motion. In particular, a mathematical process called incremental “bundle adjustment” was applied to the collected photos.
Bundle adjustment is a mathematical technique used to minimize errors that occur due to a moving camera and the resulting changing point of view during photogrammetric surveying. Once developed, the equation gives a solution to the relationship and distance between the camera and landmarks in a 3-D model. Most often this technique is used to provide more accurate information for mapping. During the 2003 MER mission, Li and his group used this technique to calculate camera positions – the location of the rovers.
Mappers At Work
One of
the first tasks encountered by Li and his research team was to
determine the precise landing sites of the rovers. A combination of
photos from the rover descent and landing site as well as satellite
images from the Mars Global Surveyor was used to accomplish this.
Prior to the landing, work had already begun on determining the precise
landing location. During the descent to Mars, a stream of photos was
collected beginning around 1,400 meters above the ground. These photos,
along with photos from the Mars Global Surveyor satellite, were used to
identify the landing sites.
“For Spirit it was relatively easy because we had a few mountains and we had a big crater,” said Li. “We could use the landmarks to triangulate the lander locations.”
Opportunity turned out to be more difficult. The rover had landed in a 22 meter crater in an area of several similar craters. Determining exactly which crater they landed in was more difficult.
“From the descent images, we knew there were quite a number of craters around Opportunity’s landing site,” said Li. “In order to determine which crater Opportunity had landed in, we built a 3-D model from the panorama images taken by Opportunity inside the crater.”
The 3-D model was measured and mapped by Li and his team. The researchers then compared the crater features, such as size and rim shape, to satellite images of the craters in the landing region to determine the exact location of Opportunity. Other techniques such as radio science, descending trajectory reconstruction, and orbital imaging were also used. The cover image shows one view of the model created by Li and his team for this task. The varying colors represent changes in terrain elevation. Opportunity’s traverse has been imposed on the image.
Over the course of the mission, Li and his team processed and analyzed thousands of images collected by the rovers, Spirit and Opportunity. This work provided interactive maps and rover location data used by the mission scientists and engineers.
“The Ohio State team created maps in near real-time, sol (a Martian day) by sol,” said Li. “We provided the maps to the mission scientists who used them to plan their scientific investigation tasks.”
Future Missions
 |
With the continuation of the 2003 Mars Mission past the planned 90 days, Li will be continuing his research along with the rest of the team; however, he is already planning the next generation of his technology. According to Li, the next potential step is to develop a real-time version that can be put on-board the rover. Programming the MarsMapper software in the rover will enable them to travel faster and over longer distances. This may allow engineers to lengthen rover operation time to several sols.
For the 2009 mission to Mars, a bigger rover traversing longer distances will be used. This continuing increase in operation area will require more advanced localization and mapping technology.
Li’s technology promises to be a valuable tool for the remote exploration of rugged environments. Not only is the technology suited for space exploration, but there could be a need for autonomous robots that will be used for mining or monitoring of hazardous environments by robots.
Contact Rongxing (Ron) Li, 614-292-6946, li.282@osu.edu