Advanced Computing in the Age of AI | Friday, March 29, 2024

Could Simulations Have Kept Mars’ Spirit Rover Rolling? 

In May 2009, the Mars rover Spirit fell through the crusty shelf of Martian topsoil to a soft bed of sand that would become its permanent resting place, despite months of desperate maneuvering by NASA engineers. But according to MIT’s Karl Iagnemma, the mission mishap could have been avoided had NASA sought the help of simulations.

When a rover gets stuck like this, the expedition’s overall goal—to collect and analyze samples throughout the planet—must be abandoned, as the rover is limited to testing only the area in which it is trapped.

Naturally, one prong of NASA’s revised approach is to avoid paths where the rover is at risk of coming upon such tricky terrain. But designing a rover that won’t be subject to the same pitfalls will offer future expeditions more flexibility in where the might explore, and the security of knowing that even in the worst case scenario, the rover won’t be lost.

To make this happen, Iagnemma says that gaining a better understanding of terramechanics—what happens when a vehicle enters deformable terrain—is in order. And while he says that we already have a solid grasp of 2,000-pound vehicles deal with this terrain, the interactions of smaller, lighter vehicles of the rover represent uncharted territory.

“There’s a lot of knowledge in civil engineering about how soils will react when subjected to heavy loads,” says Iagnemma, a principal research scientist in MIT’s Department of Mechanical Engineering. “When you take lightweight vehicles and granular soils of varying composition, it’s a very complex modeling process.”

To begin, Iagnemma and researchers from Washington University in St. Louis and the Jet Propulsion Laboratory (JPL) in Pasadena, Calif., have tested the performance of a single rover wheel as it plows its way through a tray filled with varying amounts of ultra-fine sand to gather data about torque and sinkage that will ultimately explain how the wheel develops traction in such tricky terrain.

The next step is to calibrate these data with Artemis, the team’s software model that was made specifically to simulate how the rover moves through various types of soil and terrain. To test each condition, a user plugs in a simple command, such as to drive forward for 3 feet, which mimics the instructions that the rover would receive once on Mars. The simulation then does the grunt work to determine how the rover will move over the designated soil type given the vehicle characteristics and geographical characteristics such as inclines.

Once these results were tested against the actual paths driven by previous Mars rovers, the researchers found that the simulations behaved much like the actual rovers did.

“Once you have a model you trust that is really representative of how the rover behaves, it can help mission planners make path plans in a safer way,” says team member Carmine Senatore, another research scientist at MIT. “It could say that this path looks shorter and faster, but if the soil is not what we expected, it may be much more dangerous, so it’s better to go another way.”

But you may be wondering why dry patches of soil are so much trouble in the first place - after all, when our cars get stuck it is often due to the presence of water, whether it comes in the form of mud or ice.

To answer this question, Iagnemma explains that while most of Mars’ surface is easy enough to navigate due to its firm, flat surfaces, the occasional steep dune made of ultra-fine, loose soil is the culprit.

“Think about the difference between beach sand, which you can walk on and even play volleyball on, and cake flour,” says Iagnemma. “The reason [for that difference] goes down to the microscale of the material.”

To make matters more difficult, when you get your car stuck, you have the options of getting out to push or placing a two-by-four in a tire’s path to provide it with the traction and positioning it needs to free itself. But rovers don’t have this option.

“Sometimes in a car you end up doing things like rocking it back and forth,” Iagnemma continues. “There’s limited strategies for a Mars rover because it’s not a very dynamic vehicle, and moves very slowly. So we have to be more creative and develop strategies to get out.”

To put the strategies that have emerged from Artemis’ simulations to test, the team has driven roverlike vehicles in the Mohave Desert’s Humont Dunes, and found that the vehicles’ performance was very similar to the driving patterns that the model offered. Going forward, Iagnemma says that they’d like to use Artemis to chart courses for Curiosity, which is set to navigate some of the more challenging terrain that Mars has to offer later in its mission.

“There are goals for taking the rover into places that are more difficult to travel, like dunes and steep slopes,” Iagnemma says. “That time hasn’t really been reached yet, so there’s a little time to get the model refined for Curiosity.”

Senatore, Iagnemma, Raymond Arvidson of Washington University, and their collaborators will outline additional details of the model in a paper set to appear in the Journal of Field Robotics.

EnterpriseAI