IT’S the stuff of science fiction.

The idea of intelligent robots able to think and learn for themselves has always been consigned to the imaginations of filmmakers and authors – until now.

Researchers at the University of Southampton have developed machines with artificial intelligence that can teach themselves new skills simply by reading an instruction manual.

The man behind the revolutionary new robots believes they could be a major step towards making the likes of Star Wars’ R2-D2 and C-3PO a reality.

Professor Sandor Veres has created the technology for satellites and spaceships that can react to the environment around them, spotting an asteroid or anything else whizzing through space and getting out of its way, but then also measure and record it.

He said: “And it would decide itself that it’s going to take a detour (to investigate) if it thinks something is interesting.”

Professor Veres, who is based in the university’s Autonomous Vehicle Control Systems Lab, said the technology could be used to create machines that will help on Earth, too.

He believes intelligent robots could have solved last year’s underwater BP oil spill in the Gulf of Mexico much sooner, but could also ultimately be developed to have more everyday uses by taking on laborious chores around the home.

He said: “I believe humans can be substituted in some easy tasks.

“You can buy a lawnmower, but I would expect more intelligent ways of doing more complex jobs to give some help in the garden.

“It’s a step forward towards having truly intelligent machines as we hoped for in the past.”

Most importantly, Professor Veres’ machines, which can also recognise and repair their own faults, have the ability to learn new skills, by being sent new instruction manuals.

The robots can read instructions written in plain English, thanks to his artificial intelligence system, called sysbrain.

Professor Veres said: “They can go on the web and read special documents, from which they can learn.

“We have the potential to train them and make them suitable to our needs.”

Professor Veres, who is now hoping to licence his invention, said eventually he thinks robots will have the ability to understand some subjects better than humans do.

He said: “That’s totally within the scope of our system.

“It’s very exciting. It’s been a dream for me from childhood.

“I’ve seen computers develop and there are always kind of promises of thinking machines, but nothing has happened.

“I’m not claiming we’ve got a thinking machine, but we’re going in the right direction in terms of having a dialogue between machines and human supervisors.”