As the saying goes, there will be something to be gained from the life of a foreign king. When Mars came closest to Earth, it was 55 million kilometers apart and the signal was transmitted in about 30 minutes. And as the Mars rover becomes more advanced, collecting huge amounts of data, passing them all back to Earth, waiting for human instructions to do more, it is clear that it will cost time and money. Can computers make their own on Mars?
Nasa unveiled the first results of the space probe’s next-generation intelligent system at the Goldschmidt conference on June 25, local time. They can identify the geochemical characteristics of life from rock samples and determine which data are worth analyzing and which results need to be fed back to humans.
These smart systems, while not up to the “Smanatly” rover, which will be launched in a month’s time, will make their debut in the European Space Agency’s ExoMars project. The ExoMars rover has been delayed until 2022 due to the new crown outbreak.
On Mars debut, intelligent systems will still send most of the data back to Earth. But in the future, as they move into the solar system’s more distant bodies, they will be granted autonomous decision-making by NASA.
“This is a far-sighted step in space exploration,” Victoria Da Poian, the project’s lead researcher, told the conference. This means that our thinking has gradually changed, from human involvement in all space affairs to understanding that computers are equipped with intelligent systems that can be trained to make decisions that prioritize the transmission of the most interesting and critical information. “
AI looking for life characteristics
The Rosalind Franklin Rover in the ExoMars project will carry a Mars Organic Molecular Analyzer (MOMA). This is a new mass spectrometer-based device that analyzes and identifies organic analysis in rock samples to look for existing life, or signs of past life, on the surface and underground of Mars.
The Daboan team trained artificial intelligence systems to analyze hundreds of rock samples and thousands of experimental spectra, and the first results showed that when neural network algorithms processed spectra from unknown compounds, their classification accuracy was 94% and 87% accurate lying with previously recognized samples. The algorithm will continue to be optimized until 2023.
“We get a lot of data from these unmanned missions, and sending it hundreds of millions of kilometers is challenging and extremely expensive, in other words, bandwidth is limited,” Daboan said. We need to prioritize some of the data back, while ensuring that important information is not lost. So we started developing intelligent algorithms. “
Long-term goals are self-determination
At this stage, the algorithm plays the role of auxiliary scientist, but as a long-term goal, the algorithm will independently analyze the data, adjust the debugging instrument, independently carry out the next operation, only select the most interesting data transmission home.
Eric Lynes, software director at NASA’s Goddard Space Flight Center’s Planetary Environment Laboratory, stressed that space missions face strict deadlines. Mars, for example, can store samples for a few weeks at most, then dump them and move to new locations for drilling. Sometimes, the second test of the sample is left less than 24 hours.
“In the future, when we go to explore Ganymede or Titan, the message from Earth can take five to seven hours to convey, unlike controlling a drone. We need to give these tools autonomy to make quick decisions. “
He explains that intelligent systems don’t shout ,”I’m finding life here”, but it gives scientists directions, like, “I have 87% confidence that this is phospholipids, the data is here.”