If artificial intelligence was developed to make life on Earth easier, why not use it to hunt exoplanets? So, it’s pretty reasonable to use AI for space exploration. Currently, many companies including NASA and Google have already introduced AI to the process of searching for new celestial bodies and life on other planets.
Read the article for the further information.
AI in data processing
Currently, astronomers don’t need to spend nights in front of telescopes. Modern apparatuses can observe the sky and take photos without people. However, it didn’t make the life of scientists easier.
Exploratory technologies are not limited to a couple of successful shots, they provide thousands of TB of data. For example, one telescope in Chile gives 15 TB space shots every night. Processing of such volume of information may take years if only it’s in power of people.
The point is that a telescope disfigures distant lights. It takes a long time to recognize what is illustrated on a shot: an asteroid, an exoplanet or an entire galaxy.
Bid Data analysis is a task for AI. Due to artificial intelligence, the American scientists at Stanford University and SLAC National Accelerator Laboratory reduced the time spent on hunting new space objects from months to several seconds.
The think tank used an artificial neural network to analyze half a million shots of space objects, and it demonstrated the high precision just as traditional research methods.
‘It’s amazing that artificial neural networks investigate the peculiarities to search for independently,’ says Project Head Phil Marshall. ‘It resembles kids learning to recognize objects. You do not describe a dog, just show the photos of it.’
Moreover, AI may improve the quality of photos, which has already come in handy in space exploration.
Thus, the Swiss astronaut Kevin Schawinski and his team of astrophysicists use AI to boost the resolution of blurred shots taken by telescopes. The researchers purposely lowered the quality of galaxy photos, added noise and blurring, then tested them with the help of artificial intelligence along with original pictures. As a result, the artificial neural network learned to make photos clear of the abovementioned aspects and finishing the picture.
Kevin Schawinski demonstrated shocking results, but the astronomer still takes the project with a pinch of salt.
‘After all, it’s against the principal scientific laws: we can investigate the Universe only from our personal observation,’ claims Kevin. That is why the artificial network is applied only in those areas where the results can be checked.
Hunting new space objects
Google astronomers took even a further step. By the way, how did they join this corporation?
In fact, Google introduced the ‘20% time policy’ so that the employees are free to dedicate time to their hobby one day a week. However, under the condition that this activity would bring benefit to the company. Chris Shallue and a team of AI developers made a really good use of this opportunity discovering a gap in NASA.
Googlers found out that NASA finished collecting orbital shots through the space observatory Kepler. During its four-year mission, the apparatus was taking photos every half an hour. It wasn’t possible to process all that data manually. Chris decided to help the researchers with the Google-developed AI.
As Shallue commented, it took two weeks to upload the data. All in all, the PC memory was not enough. Along with the astrophysicist and PhD at the University of Texas at Austin Andrew Vanderburg, Chris was engaged in the project for about nine months.
The enthusiasts used 15 000 NASA’s data points to create a neural network that was later trained to recognize light patterns thus detecting a planet.
The scientists didn’t manage to see some patterns at first. Later, AI discovered the exoplanets Kepler-90 and Kepler-90i that exist in a Sun-like planetary system.
Improving space technologies
Sometimes even a car is hard to drive, not even mentioning the Mars probes which one hardly ever has seen. On the other hand, sometimes technologies act before they are programmed to. For example, the Mars rover Opportunity got into a sandstorm and lost connection with us.
Taking into account this incident, NASA upgraded the software of its second Curiosity rover. The software allows the aboard camera ChemCam to independently choose observation and analysis targets.
With more freedom and independence, the survived Mars rover demonstrated more effectiveness. Nowadays, instead of waiting for the instruction from Earth, Curiosity chooses what is important to investigate further and thus collects much more useful scientific data.
As reported by NASA, it’s only the initial step towards AI integration in such technologies. A new Mars rover will be presented in 2020. It will automatically manage the process of scientific data collection taking into account own resources. Thus, the rover will uniformly distribute the amount of power with no risk of running down or losing connection with Earth.
While smart Mars rovers will be investigating Red Planet, the artificial neural network will help astronauts in open space.
In July 2018, an AI-powered robot CIMON developed by IBM and Airbus was sent to the ISS.
The absence of gravitation in open space makes any manipulations hard. It’s easy to work with CIMON since it understands a voice assistant: the crew members can ask the robot to show the needed scientific data, schemes, and instructions. This routine would not distract astronauts from their main tasks.
Even now, artificial intelligence explores completely new space areas. This leads to marvelous discoveries. Maybe, the scientists will soon find a planet that would be a home to people. But is AI powerful enough to unravel the mysteries of the Universe beginning? It remains a secret.
For more information about the possibilities of artificial intelligence, join us in Moscow at AI Conference scheduled for November 22, 2018.