“The system will basically mix the DNA of two successful parent robots to create the design for a new child robot, then print out all the parts, and assemble it completely by itself without any human in the loop at all,” explains Professor Emma Hart, chairwoman in natural computation at Edinburgh Napier University.

As part of the Autonomous Robot Evolution (ARE) project, the team has created a fully autonomous system called RoboFab that does just that.

Each of the robots it produces has a digital clone that undergoes rapid evolution in a simulated world, while its physical counterpart is tested in real-world environments.

New generations of robots are then 3D printed after combining the most successful features of a virtual “mother” and physical “father”, as well as from two virtual parents or two physical parents.

“You can imagine all of that taking place inside a box, like a factory,” says Hart.

“You could send this factory to space, and rather than having to predesign your robot and hope it works when you get there, you get the factory to design robots, build them and test them out while there. It would recycle ones that don’t work, and then regenerate new robots that are better than the previous generation.”

Cambridge University has taken a similar approach by creating a “mother” robot that can build its own “children”, test which ones do best, then modify their design. Such techniques could help with everything from the exploration of Mars to asteroid mining and the construction of lunar habitats without the need to involve humans. NASA already has a seat on the advisory board of the ARE project and has been exploring the potential of the technology.

On Earth, one of the first applications the team is exploring is the use of robots to help clean up nuclear waste in Britain. Researchers are also hoping self-replicating robots could help with search and rescue missions, as well as during deep sea dives to assemble offshore oil rigs.

However, there are major challenges. Right now, about six robots can be printed each day, featuring basic premade sensors wired into a rigid “skeleton”. The machine’s arms sometimes struggle to connect some of the sensors to the batteries with the wiring becoming tangled and requiring human intervention. Rapid advances in 3D printing, automated assembly, materials science and batteries could soon solve these issues and help create machines far more superior to any existing Mars rover.

For instance, 3D printers - once used to create low-quality plastic prototypes - are increasingly able to print using a range of materials such as metal. This would allow robots to integrate the wiring into the printing process.

Meanwhile, lighter sensors and batteries could allow for the creations of microbots small enough to crawl into tight spaces during search and rescue missions or to explore delicate features on Mars.

“What we’d really like to do is to evolve soft robots,” says Winfield. “Imagine evolving something like an octopus that literally has no bones. I mean that adds a whole other dimension of difficulty, because things that are soft are harder to control.”

However, one technological advance Winfield is wary of is artificial intelligence (AI). “We’ve already reached and gone beyond the point where we can actually understand how an AI works, and that’s a problem,” he says.

“As an engineer, you want to be able to understand what you’ve built... it might be OK in the lab, but I think it would be very dangerous to do in the real world. You never quite know what’s going to happen.”

George Zarkadakis, an AI engineer and author of In Our Own Image, believes AI is a necessary evil if robots are ever going to be able to terraform a planet for human habitation.

“How much control do we have of the robots? It’s an engineering question that should be answered. Let’s say, for instance, we send robots to the asteroid belt to mine an asteroid. What if, in the course of their evolution, they find the best way to do their work is to throw an asteroid into the path of Earth’s orbit. Could we stop it?”

Ultimately, the world will have to decide whether the benefits of creating self-replicating machines outweighs the risks. If we are ever going to live on a planet other than Earth, we may just have to depend on robot colonies to survive.

The Telegraph, London