Artificial intelligence: Helping man to explore the cosmos

Why spacecraft computing has one Hal of a future
Written by Nick Heath, Contributor

Why spacecraft computing has one Hal of a future

Man's first 50 years of spaceflight may have been fuelled by human ingenuity but it is artificial intelligence that will play an increasingly important role in future expeditions into the cosmos.

Both the European Space Agency (ESA) and Nasa are exploring how AI can be used to help satellites map other worlds and robotic rovers search for signs of life outside of Earth.

One of the first of ESA's satellite missions to use AI on its ground control systems was the Mars Express, the satellite orbiting Mars to create a high-resolution map of its surface and take measurements from the planet's atmosphere and surface.

Since 2005, the AI software has been helping mission planners schedule the best time each day for the Mars Express to dump the data it has gathered to the ESA systems back on Earth, ensuring the data is not lost in transit or overwritten in the satellite's memory.

Calculating the best time to download data requires the AI software to take a host of factors into account, such as the orientation of the spacecraft, the planned scientific activity, the ground station availability and the bandwidth available for space-ground communication.

"Before, this was very tedious and very complicated manual work - with the [AI] tool we reduced more than 50 per cent of the workload associated with this task," Alessandro Donati, head of the advanced mission concepts and technologies office at ESA's Space Operations Centre at Darmstadt, Germany, told silicon.com.

The surface of Mars taken by the Mars Express orbiter in 2007. It shows the Hephaestus Fossae region, which is dotted with craters and channel systems

An image of the surface of Mars taken by the Mars Express orbiter in 2007
(Photo credit: Nasa)

AI-based scheduling is also used by ESA ground staff to help plan where in the universe the Integral gamma ray space telescope should collect data from.

Previously it would have solely been up to a member of the ground staff to schedule Integral's data-gathering activities. "This is a fantastic improvement in terms of workload," Donati said.

ESA is also developing AI software that will enable planet exploration rovers, and eventually satellites, to automatically reschedule their duties - a useful ability for when an unexpected event or discovery upsets its normal routine.

"It is a very important building block towards the onboard autonomy process," said Donati.

"The software takes into consideration events that have happened and availability of resources that might not necessarily be the same as we were expecting at the beginning of the process."

This "autonomous controller" software is currently being developed within ESA and the next stage of its development will be to test it using a software simulation of a real rover mission.

The first satellite missions to use these more sophisticated AI programs in onboard systems are likely to either be...

...smaller scale missions launched by European universities or ESA's own Proba craft - the mini satellites the space agency uses to show technologies can work in orbit before they are used on board commercial or research satellites.

Installing more sophisticated AI routines on board satellites and rovers will allow them to carry out tasks that are not possible today.

Donati cites the example of the Earth observation satellites, which could use AI software to begin taking pictures and emission measurements from a volcano as soon as it starts erupting.

"You could think of how this could have been used to monitor the eruption of the volcano in Iceland," Donati said.

"Using data from its infrared sensors the system can recognise a spot that represents a huge amount of heat, and work out that this is most likely a volcano.

"Then, instead of a person having to program the satellite and ask it to take the first pictures of the eruption, the satellite is autonomously able to reprogram its planned activities and identify the right time when to take these pictures and then dump them to ground. All you need is to give the satellites its goals."

The Nasa Earth observation satellite EO-1 has already demonstrated the use of AI to help spot an erupting volcano in Indonesia

The Nasa Earth observation satellite EO-1 has already demonstrated the use of AI to help spot an erupting volcano
(Photo credit: NASA Goddard/Reto Stockli and Robert Simmon)

The Nasa Earth observation satellite EO-1 has already demonstrated the use of AI to make these kind of decisions: in 2005, EO-1 spotted the plume of ash from an erupting volcano in Indonesia and immediately began collecting data from the volcano.

As well as using data collected by its own onboard sensors, EO-1 is also able to draw on data from other satellites and ground stations to decide what it should do next.

But it is in the field of deep space exploration where onboard AI could really revolutionise what spacecraft are able to do.

Because deep space probes are so far away from the Earth and it takes so long for ground commands to reach the spacecraft, ground staff have to queue up orders for the probe ahead of time, scheduling when the spacecraft should carry out specific commands.

"In low-Earth-orbit satellites, you can do command and control and monitor the effect in near real time. In exploration missions it's not possible - it has to all be plotted out in advance," said Donati.

"That's why AI is helpful because you need something more advanced... to allow the spacecraft to intelligently dump existing priorities and reschedule tasks on the go."

AI will also help rovers be more efficient when exploring alien planets Donati added.

"When they find a rock that is of particular interest they can decide on their own to...

...take a snapshot and reprogram their other planned activities - without having to wait for new commands to be sent the following day telling them what to do."

With AI set to play a greater role on board spacecraft and rovers in future, ESA is also developing intelligent planning software that will help the agency's ground staff to support missions where a rover or spacecraft has a high degree of onboard autonomy.

"The planning and scheduling from ground of a goal-based autonomous element is totally different," said Donati.

"From the ground you plan the goals but you no longer have to plan the nitty gritty of the activities - those goals are digested and refined onboard by the software."

As spacecraft and rovers begin to make more decisions for themselves and software is used to automate tasks within mission control hubs on the ground, future space missions will be able to reduce the numbers of staff and equipment needed to carry out routine tasks.

Artificial intelligence in space: ESA's ExoMars rover will be capable of operating autonomously during its mission to look for signs of past or present life on Mars

An artist's impression of ESA's ExoMars rover: it will be capable of operating autonomously during its mission to look for signs of past or present life on Mars
(Image credit: ESA)

"Certain things that are done on ground will be able to be done easily onboard," Donati said.

The consequence of installing AI routines on board spacecraft and rovers, however, is that the software will both be more complex, and therefore more likely to go wrong, requiring more staff on the ground who specialise in diagnosing and fixing problems in complex computer systems.

"Any time you increase the complexity on board [the spacecraft or rover] you increase the risk because there's more to go wrong when there's more software running," Donati said.

"This is why such sophisticated onboard software will be used in the future where it is strictly needed and justified."

As exciting as the possibility of fully autonomous spacecraft are, space agencies' cautious approach to adopting new technologies means the adoption of onboard AI software is likely to take time.

Given the logistical difficulty of fixing spacecraft and the high cost of failure, Donati said that space agencies will always plump for tried and tested tech over the cutting edge.

"The environment is very risk-conscious - you are always faced with the fact that if a piece of hardware doesn't work, you can't just say 'Let's replace it' - you simply cannot," Donati said.

It may be slow tech but it's a tried and tested approach that has carried man to the moon and an unmanned probe beyond the boundaries of our solar system - who knows where it could take us over the next 50 years?

If you like this story then why not check out other articles on the technology powering man's exploration of space:

Editorial standards