Artificial Intelligence in Space

“I am putting myself to the fullest possible use, which is all I think that any conscious entity can ever hope to do.” – HAL, 2001: A Space Odyssey

Artificial intelligence in space sci-fi is both the stuff of dreams and nightmares. In 1968 the world was introduced to HAL 9000, a sentient computer onboard the Discovery One spacecraft in Arthur C Clarke’s 2001: A Space Odyssey. Designed to maintain the spacecraft and help the crew with tasks on their journey to Jupiter, over time HAL begins to malfunction and takes drastic action to prevent the crew from disconnecting it.

This chilling tale explores the extreme scenario of a machine that can think, reason, and learn at a high level. Today, with the year 2001 long gone, we’ve still yet to develop a machine that even comes close to HAL’s level of intelligence. But there are many good reasons why some form of artificial intelligence (or AI) is crucial for future space exploration. Let’s just hope we never have to ask it to ‘open the pod bay door’!

What is AI?

What do we mean by artificial intelligence anyway? Are our computers and phones intelligent because they can correct spelling mistakes and calculate the solutions to complex mathematical equations? No. These computers are just doing what we’ve pre-programmed them to do, at speeds much faster than any human.

Artificial intelligence is the ability of a machine to perceive and recognise its environment, learn from previous experiences, and make logical decisions to new questions and scenarios towards some end goal.

Being intelligent is like the difference between baking a cake following a step-by-step recipe and improvising a new dessert based on extensive previous experience with ingredients, baking methods, and preferences of customers. Any computer code can tell a robot how to follow a cake recipe. But an AI machine could create a new cake… and it would be tasty too.

Over the past two decades, we’ve seen computers beat professionals at chess (IBM’s Deep Blue in 1998), Jeopardy! (IBM’s Watson in 2011), Go (DeepMind’s AlphaGo in 2016), and, very recently, even at poker (Carnegie Mellon University’s Liberatus in January 2017). Computers can now routinely recognise human speech, handwriting, and images with ever-improving accuracy.

How can it be used in space?

The remote, hostile environment of space is a natural fit for AI machines, for three key reasons:

1. Space is too vast for rapid communication back to Earth. When we talk to astronauts or computers in orbit around the Earth the time delay is small – less than a second for a radio signal to travel up to the ISS and back down to Earth. But it takes light 2.6 seconds to travel to the Moon and back, 6-42 minutes to travel to Mars and back, and more than an hour to travel to Jupiter and back! As our spacecraft and astronauts venture further out into the Solar System, it simply isn’t possible to relay the commands for every action from Earth. We need on-board computers to think for themselves.

2. Spacecraft and mission schedules are incredibly complex systems, relying on countless small mechanics and actions in order to achieve some goal. Planning the daily schedule for a remote space mission such as a rover on Mars can take a whole team of people hours each day. But with the help of AI, humans can tell computers the big-picture goal (such as ‘travel to that crater’) and the rover’s computers can plan the detailed series of steps needed to make it happen, including dealing with any minor problems that arise. Back in 1998, NASA successfully tested an AI-like software called Remote Agent on the Deep Space 1 spacecraft. Remote Agent had the ability to both plan a schedule and diagnose with faults on the spacecraft, all without human interference. Today, AI software developed from Remote Agent is used by Martian rovers Opportunity and Curiosity as well as satellites such as Earth Observing-1 (an Earth Observation pioneer that sadly reaches the end of its mission on 30 March 2017 after 17 years).

3. Space is hostile to human life and in some cases impossible for humans to explore. The moons of Jupiter, in particular Ganymede and Europa, are a compelling place to hunt for signs of life. We believe they harbour vast liquid water oceans beneath their icy crusts, and perhaps where there is water, there is microbial, alien life. But the intense radiation fields around Jupiter are simply inhospitable for human life. To search for life on these moons, we would have to send a robotic mission in our stead. And with more than an hour time delay in communications between Earth and Jupiter, that robotic mission would have to be able to search for and even recognise life autonomously, without our help.

Closer to home, NASA is developing a robot called Robonaut for the International Space Station. Eventually, the hope is that Robonaut will perform risky spacewalks while astronauts guide it from the safety of inside the space station.

Space exploration is a complex, hostile endeavour that takes place over vast distances from Earth and artificial intelligence is key to helping humans both on Earth and in space make this exploration feasible.

We’re a long way off a completely autonomous AI system such as HAL, but our satellites around Earth and rovers on Mars already rely on AI-like software to plan their activities and select the most interesting features to study.

For all future robotic and manned missions beyond low-Earth orbit, you can bet that AI will be our constant companion.