The March 2020 rover is likely to carry artificial intelligence that helps handle the workload for science.  Welcome to Ars UNITE, our week-long virtual conference on how innovations lead to unusual pairings. Every day of the week from Wednesday to Friday, we will get a couple of stories about the future. Today’s focus is on AI in manufacturing and space racks by blasting!
NASA can not yet put a researcher on Mars. But in its next rover mission to the Red Planet, NASA’s Jet Propulsion Laboratory hopes to use artificial intelligence to at least add a talented research assistant there. Steve Chien, director of the AI Group on NASA JPL, intends to work with Mars 2020 Rover “much more like [how]” you would interact with a doctoral student instead of a rover you usually need to micronize. “
The 13-minute delay in communication between Earth and Mars means that the movements and experiments made by former and present Martians have had to plan. While the last robbers have the ability to recognize hazards and perform certain tasks autonomously, they still have high demands on their support teams.
Chien sees AI’s future role in the human space flow program as one where people focus on hard parts, like steering robots in a natural way while the machines work autonomously and give people a high level summary.
“I’ll be almost like a partner with us,” predicted Chien. “It will try, and then we say” No, try something longer, because I think it looks better “and then try. It understands what elongated means and it knows a lot of details, like trying to fly the formations
“However, it is obviously on the dystopian level,” Chien joke. But he does not see it happening soon. “
Old School Autonomy
NASA has a long history of AI and machine learning technology “says Chien. Much of that story has been focused on using machine learning to interpret extremely large amounts of data. Many of the machine learning involved spacecraft data sent back to earth for processing, there’s a good reason to add more intelligence directly to spacecraft: to handle the communications volume.
Earth Observing One was an early example of deploying intelligence aboard a spacecraft. Launched in November 2 000 EO-1 was originally scheduled to have a year’s mission, some of which were to test how basic AI could handle certain scientific data on board. One of the AI systems tested on the EO-1 was Autonome Sciencecraft Experiment (ASE), a set of software that enabled the satellite to make decisions based on data collected by its image sensors. ASE included on board science algorithms that performed image data analysis to detect triggering conditions to make spacecraft more aware of something, such as interesting features detected or modified in relation to previous observations. The software can also detect cloud cover and edit it from final image packages transferred at home. EO-1 ASE can also customize satellite activities based on science collected in a previous circulation.
With the volcanic images, for example, said Chien, JPL had trained the machine learning system to recognize volcanic eruptions from spectral and image data. When the software detected an outbreak, it would then work preprogrammed policies on how to use that data and schedule follow-up observations. For example, researchers can set the following policy: If the spacecraft exhibits a thermal emission exceeding two megawatts, the spacecraft should keep it observable at the next overflow. The AI software aboard the spacecraft already knows when it will overflow the release next, so it calculates the amount of space required for the observation on the solid state recorder and all other variables required for the next passport. The software can also drive other observations for a circulation to prioritize emerging science.
2020 and beyond
“It’s a good example of things we could do and now pushed forward in the future to more complicated missions,” said Chien. “Now, we are looking to put a similar scheduling system onboard the Mars 2020 rover, which is much more complicated. Since a satellite follows a highly predictable orbital, the only variable an orbiter has to handle scientific data collecting.
” When you plan to take a picture of this volcano at 10 o’clock, you’ll take quite a picture of the volcano at 10 o’clock, because it’s very easy to predict, “continued Chien.” What is unpredictable is whether the volcano erupted or not, so AI used to answer it. “A rover, on the other hand, has to handle a large set of environment variables that change moments at a moment.
Even for a orbit satellite, scheduling observations can be very complicated. So, AI plays an important role even when a person makes the decisions, says Chien.” Depending on mission complexity and how many limitations you can get into the software, it can happen automatically or with AI that increases the capabilities of the person’s capabilities. The person can fish with priorities and see what different schedules are coming out and explore a larger part of the space to reach better plans. For easier assignments, we can only automate it. “
Despite the lessons from EO-1, Chien said that spacecraft with AI remains” the exception, not the norm. ” I can tell you about different space assignments using AI, but if you chose a space assignment randomly, the chances of using AI in any essential way are very low. As practitioners, there is something we need to get busy with. It will be a big change. “