How NASA Is Helping Humans Reach the Red Planet, Using GPUs and 3D Visualization

by Isha Salian

It can take up to 10 months for a spacecraft to get from Earth to Mars. But the entire journey can be in vain if something goes wrong in the last six minutes.

That’s how long the vehicle has to slow from its initial descent rate of 12,000 miles per hour to near zero for a smooth landing. The process is incredibly complicated, and the lander will only have one shot to get it right.

“Landing’s not something you can bail out on,” says Ashley Korzun, a researcher in the Atmospheric Flight and Entry Systems Branch at the NASA Langley Research Center. “Once you start, you’re going to the surface.”

To plan the landing for NASA’s first manned mission to Mars, the agency is relying on high-resolution, NVIDIA GPU-powered simulations of the complex physics involved.

Onstage at the SC19 supercomputing show in Denver, NVIDIA founder and CEO Jensen Huang wowed audiences with a demo created using NVIDIA IndeX volume visualization software — displaying a stunning visualization of the velocity field around a lander as it hurtles towards the Martian surface.

Simulations this detailed and responsive are unprecedented, Korzun said. “The ability to move your way around the solution in real time is something we’ve never been able to do, even with small datasets. To be able to do that with the highest-resolution datasets we’ve gathered is truly groundbreaking.”

SC19 attendees can try out the interactive Mars lander visualization for themselves in NVIDIA booth 901, running on a cluster of NVIDIA DGX-2 systems. The underlying dataset is a 6-billion cell, unstructured mesh of tetrahedra, prisms and pyramids, making it the world’s largest interactive volume visualization.

The data behind the simulation was crunched on the fastest supercomputer in the world, the NVIDIA GPU-powered Summit system at Oak Ridge National Laboratory.

Sticking the Mars Landing

While NASA has sent humans to the moon and rovers to Mars, a human mission to the Red Planet is more than a decade away. Before a mission can launch in the 2030s, there’s a major challenge to solve: how to safely land heavy loads on the planet.

Past Mars trips have used landers that weigh around a ton, allowing space agencies to rely on air resistance and a parachute to safely get a vehicle through the atmosphere and to the Martian surface.

“That paradigm completely breaks down when we go to the vehicle size needed to support human missions,” said Eric Nielsen, a senior research scientist at NASA Langley.

A manned vehicle will weigh tens of metric tons, at least. That’s 10x more than the Curiosity rover, NASA’s heaviest lander to reach the planet, which weighed about as much as a compact car. Mars’ atmosphere will absorb most of the vehicle’s kinetic energy as it hurtles down to the surface, but not nearly enough for a soft landing.

Scaling Up Simulations on Summit

NASA scientists and engineers are working with colleagues from Old Dominion University and NVIDIA to simulate a solution that uses retropropulsion. Firing a lander’s engines in the opposite direction of the planet’s surface at supersonic conditions will create an upward thrust that can slow down the heavy vehicle’s descent. Complex fluid dynamics will be at work, from tiny turbulent eddies just centimeters from the vehicle to larger airflow meters away.

That’s why NASA relies on high-resolution simulations to plan how to control the speed and orientation of the vehicle under different landing conditions. Each simulation produces over 100 terabytes of output data, Nielsen said. Hundreds of thousands of them will be run to test a range of possible conditions a vehicle could encounter during its descent to the Martian surface.

“If we were to run these studies on a conventional computing platform operating in a capacity environment, one simulation would require up to six months,” said Aaron Walden, a research computer scientist at NASA Langley, in a recent talk at GTC DC — the Washington edition of NVIDIA’s GPU Technology Conference. “On Summit, that simulation takes about a workweek to run, and we can actually perform six such jobs simultaneously.”

Using the processing power of 3,312 NVIDIA V100 Tensor Core GPUs, the team can run an ensemble of six simulations at once with NASA’s FUN3D computational fluid dynamics software. Running simulations on Summit also allows the team to incorporate a much higher resolution than prior projects, running large-scale problems while still capturing more of the physics involved than ever before.

“It’s been pretty game-changing in terms of the learning cycle,” Nielsen said.

Visualizing the Last-Mile Problem

Each FUN3D simulation shows a specific moment in time along the lander’s descent trajectory. After being processed on Summit, each one can be rendered into a visualization showing different fluid dynamics variables such as density, vorticity and velocity.

Using the NVIDIA IndeX 3D volumetric visualization SDK, a dynamic simulation can for the first time be generated out of the FUN3D data, each simulation measuring a colossal 128 terabytes (as much as over 1,000 4K movies). The complex rendering runs on a cluster of NVIDIA DGX-2 systems, using GPUDirect Storage technology for high-bandwidth data transfers.

The visualization allows users to interact with the results in real time by changing the viewing angle, or zooming in and out to examine what’s happening right at the rocket engine nozzle exit, as well as tens of meters ahead.

“Being able to interrogate solutions in that way adds another dimension that’s incredibly enabling to our designing these vehicles, especially in a collaborative environment,” Korzun said.

Gaining a more detailed picture of the force fields on the vehicle when flying with retropropulsion can inform the team’s engine design choices for the Mars lander. They could use the visualizations to detect necessary adjustments, find design optimizations and test different configurations of how the engines integrate with the vehicle.

Main image produced by NVIDIA IndeX visualization technology.