Big data, AI, and Space exploration
Space is huge. Some even believe it could be infinite. Since space is so huge, studying it produces a massive amount of data. Big data and AI is coming together to help answer some of the greatest unanswered questions in cosmology and space exploration.
The role of AI in space exploration is becoming increasingly vital. Recently a group of physicists, mathematicians, data scientists, and engineers was formed to take the first ever picture of the supermassive black hole, and machine learning plays an important role in the project.
Supermassive Black Hole
A supermassive black hole appears to inhabit the center of the Milky Way galaxy, about 25,000 light-years away. Black holes suck in light, which makes them invisible. They can only be observed by looking for their immense gravitational pull. This pull produces a ring-like structure around the black hole. As the light approaches the event horizon of a black hole, the point beyond which it cannot escape, it bends and stretches. But the ring is very small, and due to diffraction, there are fundamental limits to the smallest objects that can be observed. This means in order to see smaller and smaller cosmic objects, telescopes need to get bigger and bigger. To view the event horizon around the supermassive black hole at the center of our galaxy a telescope the size of the earth would need to be built. That, of course, is impossible.
However, there is a way in which this limitation can be overcome. The alternative solution is to connect telescopes from around the world, in an international collaboration called the Event Horizon Telescope (EHT). The EHT creates joins together telescopes and so creates a virtual telescope array the size of the earth. This is capable of seeing structures on the scale of a black holes event horizon.
Dr. Katie Bouman, Ph.D., from the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT is one of the members of the Massachusetts-based group. Her role in the endeavor is pivotal. Each telescope in the worldwide network works together, linked by the precise timing of atomic clocks. Teams of researchers at each of the sight collect thousands of terabytes of data. This data is then processed in a lab in Massachusetts.
So how exactly does machine learning aid this project? Individual telescopes receive only a part of the light from the ring wrapped around the event horizon. This photo misses many parts of the image. By using a deep learning algorithm the missing parts of the image are constructed.
AIs Exploring Mars
Rovers on Mars are using investigating whether the red planet once harbored life and if it could support a human colony. They cannot be controlled by a human operator since it takes between three and twenty minutes for a signal to reach Mars. So if the rover ready to drive off a cliff, it would to too late to send the signal. Rovers have to be able to operate autonomously, and this requires some type of intelligence.
Many companies, like Spire, SpaceKnow, Planet Labs, DigitalGlobe, and SpaceX are making millions of dollars from selling data from satellite observations and imagery. There are many potential applications of earth observation data. Monitoring weather, agricultural purposes, measuring gravitational differences across various land formations, and calculating the concentration of the different elements in specific areas, among others. A learning algorithm must manipulate a huge amount of data if all these potential applications are going to become reality.
The Future of the Space Industry
Companies are willing to pay a lot for this kind of information. This data would be useful in a large number of sectors such as hydrocarbons, agriculture, mining, humanitarian, shipping, security, and national security. However, the raw data is complex and often mind-boggling to a human, and it has to be wrapped up and presented in a way so that a client can understand and use it. Space exploration and exploitation has a bright future, and AI is a central part of it.
