Recalculating: NASA Developing GPS for Space Using Artificial Intelligence

Every good space crew needs a great astrogator. Charting a course through stars and debris fields is a lot harder than driving through a city because there is no infrastructure to use as a guide, but NASA is working to solve that problem. 

In the history of space travel, astronauts have operated on a Point A to Point B system. Staying on course is important, but it's not like they are up there turning giant steering wheels and checking the direction of the wind like space pirates. To prevent future astronauts and space tourists from getting lost as trips out of our atmosphere become more frequent, a team at NASA's Frontier Development Lab (FDL) is developing a sort of "non-global" GPS system that uses artificial intelligence to give directions. 

According to Futurism, a team at FDL found that an AI could identify a location if it were first taught using photographs. By showing the program enough photos of the surface of moon or exoplanet, the team would be training it to build a virtual model, so when a lost traveler took a photo of his/her surroundings, the AI would cross-reference it with the virtual map and be able to give directions to a desired point. 

To test the system, the team developed a system to simulation the surface of the moon and provided the AI with 2.4 million images of it. During a presentation to conclude FDL's summer program, the team reported that the test was successful, and revealed that the next step would be to use actual photos of Mars-not merely a simulation of the planet- to see how the artificial intelligence would do. In science fiction, characters sometimes use probes to fly around and create three-dimensional models of a space, which is kind of what the FDL team is doing in a more plausible (by 2018 standards) way. This kind of system would make exploration and colonization of other planets a lot easier, but we suspect that changing landforms and visibility conditions would make it harder for the AI to recognize locations solely based on images.