Could Genetic Algorithms Boost Space Probe Intelligence?

Voyager carried out a gravitational slingshot manoeuvre past Jupiter (NASA)

Rocket science ain’t easy, but what about celestial navigation? Once you’ve launched your probe into space, surely the hard bit has been done, and we can sit back and relax, happy in the knowledge our technology is out of the Earth’s hefty gravitational well? The robot is coasting through the vacuum of space ready to accomplish some science. Job done. Easy.

As you may have guessed, it isn’t that easy, in fact sending a spaceship on the equivalent of a Solar System-scale game of gravitational ping-pong is highly problematic. What if your launch is delayed? What if the inter-planetary medium (the stuff between the planets) is of a higher density than you expected? Perhaps the Sun has pumped out more particles than you had calculated pre-launch, creating drag and slowing your spaceship down?

Unfortunately, once the spacecraft is on its way, apart from a few minor Earth-commanded corrections allowed by the ship’s thrusters (wasting valuable fuel), the spaceship is by itself, hoping your calculations are as complete as they can be.

When the spaceship in question has to use planetary gravity assists to accelerate or decelerate on its journey to a deep space destination, slight deviations in trajectory than what was calculated can result in inefficient sling-shots or even complete loss of the mission.

Now Ian Carnelli and colleagues from ESA in Noordwijk (Holland) have prepared a publication that details a possible solution using a genetic algorithm. Basically, the computer on board a next generation space probe could simulate multiple autopilots guiding a virtual version of the probe. Each autopilot executes its code and the computer will select which simulated autopilot performs the best (i.e. solutions that waste fuel or find the slowest route will be ignored).

Happy with the best group of simulated solutions, the computer will selectively “breed” them together to develop an optimized pilot, with no need to wait for instructions to be sent from Earth. “After hundreds of generations of the GA you obtain a ‘pilot’ that is an extremely good performer – able to fly the assist trajectory that uses the least propellant while reaching the next target planet faster,” Carnelli says.

Using simulations here on Earth, Carnelli has successfully used his genetic algorithm to optimize the trip of a virtual spaceship to Pluto via Jupiter and another to Mercury via Venus.

Although installing this system on missions in the near future may not be a possibility, it is a tantalizing look into how unnatural selection could be used to optimize, and therefore protecting, expensive pieces of kit in deep space.

Source: New Scientist

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: