The team is still "training" the software that will control the bots to "recognise" corals and distinguish them from other sea objects.
Corals are easily damaged by pollution and destructive fishing practices, and it takes decades for them to re-grow.
When they get damaged, scuba divers re-cement broken fragments, helping them re-grow - but it is tricky for divers to reach depths over 200m.
Coralbots, the researchers hope, will be a lot more efficient, able to repair the reefs in days or weeks.
The team, which consists of a marine biologist, an artificial intelligence scientist, a roboticist, and a machine vision scientist, said it was trying to raise £2m to hold a first demonstration.
The scientists said that if they got all the cash they needed, the bots could be embark on their first mission within a year.
Initially, the robots would be adaptations of those already developed at the university's Ocean Systems Lab.
They would be about a metre long, with built-in video, image-processing and simple manipulation tools, such as scoops and arms, and would operate in "swarms".
“This project explores one of the most intriguing and impressive feats of natural swarm intelligence, whereby collections of simple-minded individuals collaborate to construct complex and functional structures,” says Heriot-Watt Professor David Corne. “Exactly how this happens is only partly understood, but scientists have several clues and ideas, and we will exploit these ideas to achieve reef reconstruction.”
Swarm robots have an added benefit to the project in that they reduce the engineering requirements for extremely robust robots. If one coralbot is damaged, there are many more that can complete the task. The robots are also easily deployed in emergency situations, such as when a hurricane or trawling damages a nearby reef.