Artificial Intelligence (AI) is expected to change the way NASA studies space. NASA Earth Science Satellite equipped with AI to detect the 2010 Icelandic volcanic eruptions to help generate this color picture.
How to make a robot to identify unexpected events? That is the problem that artificial intelligence researchers think about, especially when AI starts to change space research.
“Science: Robotics” (Science: Robotics) magazine published a new article, briefly describes how artificial intelligence in the space mission to explore activities. This article is co-authored by Steve Chien and Kiri Wagstaff of NASA JPL, Pasadena, California. The reference to autonomous control will be the key to exploring the future of the solar system, because in the solar system, the robot spacecraft and ground control personnel often lost. In a sense, space scientists are in the robot spacecraft with the help of field visits.
“Our goal is to make AI more like a smart assistant with scientists, not programming code,” said Chien, a senior research scientist at the autonomous space system. “This allows scientists to focus on ‘thinking’ – analyzing and parsing data – and robotic detectors focus on searching for research-related features. ”
Science is driven by an unknown thing, and it is easier for a trained person to find things beyond what is expected. But for robots, this means having the ability to judge whether the situation is “normal” and the ability to detect statistical anomalies using machine learning techniques.
“We do not want to miss because I do not know how to find it,” said Wagstaff, chief data scientist at the JPL Machine Learning Group. “We want the spacecraft to know what we expect to see and identify it when we see it.”
Detection of abnormal features is one of the functions of AI. But there is a more complex function will be in the study of the marine world (such as Europa) in the side.
“If your background is so rich, you can build a normal model that the robot can expect,” Wagstaff said. “But for the new environment, we want the spacecraft to build a normal model based on their own observations.” Can identify the unexpected situation we can not expect. ”
“Imagine, for example, that AI locates the eruption of the seabed, which is sudden and ever-changing, and that allows a passing spacecraft to re-optimize the order of the task and instantly study the phenomenon,” Chien said.
JPL is mainly responsible for the development of several key cases of space AI. The dust whirlwind on the surface of Mars was taken by the NASA Opportunity Mars “WATCH” program. This program was later developed into AEGIS to help curiosity ChemCam chose a new laser target that meets both the scientific team’s parameters and the need to wait for interaction with scientists on Earth. AEGIS can also fine tune ChemCam’s laser pointing.
The Earth Observing-1 (EO-1) satellite, managed by the NASA Goddard Space Flight Center in Greenbelt, Maryland, is equipped with a satellite called the Autonomous Sciencecraft Experiment, AI software has studied volcanoes, floods and fires. The EO-1 Titan equipment also uses AI to identify glacier deposits on the surface of glaciers, which is important for places such as Europa, since sulfur deposits are likely to imply potential life characteristics.
AI allows spacecraft to process the collected data by level to achieve a balance between demand for power supply and limited data storage. The self-management of this system will be applied to the NASA 2020 rover, and the AEGIS program will also be used to select the laser target.
While the autonomy has brought exciting new advantages to the scientific team, both Chien and Wagstaff have stressed that AI still has a long way to go.
“In the foreseeable future, high-level talent is still the mainstay of scientific development,” Wagstaff said: “But artificial intelligence is an observation tool, and can use a unique way to assist human research science.”