Machine Learning: Now, Then and in the Future
Machine learning, artificial intelligence (AI), autonomous systems – three phrases that all stand for computers making decisions independently. By utilizing algorithms and a lot of data, computers are increasingly capable of performing tasks without needing permission from a human operator. Machine learning technology is gaining popularity, especially as mega constellations and human exploration missions beyond low Earth orbit (LEO) are becoming a reality.
History of machine learning in the space industry so far
Machine learning is not completely new: it has been used in previous missions including deep space exploration, satellites and planet exploration. One of these was the Deep Space 1 (DS1) technology demonstration mission, which was launched in 1998 and aimed to fly by a comet and an asteroid. The spacecraft carried a remote, intelligent self-repair software system, called Remote Agent Experiment, also known as RAX, which was developed by JPL and NASA Ames. This system is known as the first artificially intelligent system used on a spacecraft. Some of the software and components that were part of this spacecraft have been used on other NASA missions.
In 2000, NASA Goddard Space Flight Center launched the Earth Observing-1 (EO-1) observation satellite on a Delta II rocket. This satellite was built to enhance development of future Earth observing satellites. It was fitted with the Autonomous Sciencecraft Experiment software, which was used as a test on the EO-1 and allowed the satellite to capture images autonomously. The software was supported with continuous cloud cover updates from NOAA as well as a priority list of images.
Other missions that use autonomous software are the Mars rovers Spirit and Opportunity. Both have received several software upgrades since their launch in 2003. Opportunity received an upgrade that included Autonomous Exploration for Gathering Increased Science (AEGIS) software, which allows the rover to analyse images with its own camera to further examine specific rocks that meet certain criteria like color and shape. The software takes away the time-consuming step of the rover transmitting the images back to Earth to be analyzed by operators, who would then transmit data back to the Rover to tell it what to do next.
Future of exploration missions
Several major players in the space industry have set their sights on (manned) exploration missions beyond low Earth orbit. To establish such missions, the industry will need to rely on extremely smart artificial intelligence technology to support astronauts and those back on Earth during those missions. AI would be used for rendezvous and docking to with space stations. It would also be used for sample selection for in-situ resource utilization on the Moon or other celestial bodies, to guarantee supplies for crew and spacecraft far away, for example water for the use of crew, and also for using this water to make propellant.
An example of AI assisting astronauts is the Crew Interactive Mobile Companion (CIMON), which was developed by Airbus Defence & Space for the German space agency DLR. CIMON is an experiment and will make its way to the International Space Station in the summer of 2018, when German astronaut Alexander Gerst starts his mission on the space station. As stated by Airbus, CIMON is ‘designed to assist and to create a feeling of talking to a crew mate’, who also happens to be very smart!
On the other side of the Atlantic, Boeing and SpaceX are integrating AI into their CST-100 Starliner and Dragon capsules. Having an ‘automatic pilot’ on board those vehicles reduces some of the crew’s comprehensive training schedule, said Chris Ferguson at Boeing, because they will only have to learn the important things, such as what to do if the spacecraft does something it should not do.
Meanwhile, the technology is also being further developed for unmanned exploration missions. Opportunity is still happily exploring Mars, and was joined by its cousin rover Curiosity in 2012 to explore the area of the Gale Crater. In 2016, Curiosity received an upgrade to AEGIS. While having the same analysis capabilities to understand the size and height of a rock, Curiosity is also able to examine the rock further by using its ‘ChemCam’ – a laser that vaporizes the rock so that the rover can analyze its composition.
There is much in store for AI technology in the space industry in the coming years. As our missions grow bolder and more ambitious, this technology will be increasingly referred to.
Learn more about the future of machine learning on Thursday May 24 at the Open Tech Forum at Space Tech Expo 2018.