Swarms of beetle-like robots using advanced sensory motor control technology could one day be used to crawl over the rubble of disaster areas in search of survivors.
The rescue robots are just one of the possible outcomes of a three-year research project at Edinburgh University, where a team of engineers is attempting to create a six-legged robot that can adapt its gait to complex terrain.
'Getting a robot to walk on flat territory is a well-solved problem,' said Alan Murray of the university's Neural Networks Group. 'When it gets bumpy it becomes difficult.'
Murray, with project leader Barbara Webb from the university's Insect Robotics Group, will design a robot that uses vision and tactile sensors to navigate homes, buildings and the outdoors. The robot will be equipped with a small camera and a vision algorithm that will allow it to see, recognise and avoid running into objects.
'It will be able to look for edges in an image,' Murray said. 'For example, it would be able to figure out when there's a hole in the ground.'
The robot will use its tactile sensors to determine where its limbs are located and will rely on an artificial neural network to direct its appendages. For example, if it did come upon a hole, it could decide whether to walk around it, stretch its legs and step over it or walk into and out of it.
'There are many other research groups building insect-like robots, but the robots tend to be ones that do only reflex behaviours,' said Webb. 'There is not much research on building insect-like robots that continually learn what to do.'
There are many examples of humanoid robots that display these sort of learning mechanisms. One popular example is the Honda Motor Company's ASIMO, which has already been proven to walk up complex terrain such as stairs.
The hexapod robot, however, would be smaller and cheaper and therefore more suited for dangerous environments such as the rubble ruins of a destroyed building.
'In those situations it would be better to send lots of simple, cheap robots rather than one big expensive one,' said Webb.
Theoretically, it should be easier to model a robot on an insect than on a human. One reason is that a human brain is much more complex than an insect brain.
Murray, who is an electrical engineer, has spent years researching ways to build computer chips that work like the brain.
'In the brain neurons send messages to one another in a sort of neural Morse code,' he said. 'They don't send digital bits like computers, and they don't send analogue data like we find in audio systems. They communicate via little blips of signal — voltage spikes.'
Murray will find out how a robot can use these spikes in voltage to detect edges and the movement of edges in an image in order to work out how far it is from something.
The Edinburgh team decided to build a robot with six legs because that provides stability with manageable control challenges. Each leg would contain three motors.
'While eight legs would arguably be more stable than six legs, it would be just more to control to no avail,' Murray said.
The robot would measure about 30-50cm long and the legs would be 20cm high. However, if the research team is able to prove the concept, the size of the robot will be scalable.
'If you're sending something into a rubble-strewn area, the bigger it is the bigger obstacles it can cope with,' Murray said.
The robot could also be used in much less disastrous situations. Murray suggested that the six-legged machine could be an ideal household robot.
'Unlike a robot on wheels, this robot could go up the stairs,' he said.
'If people could cope with the idea of something that looked like a big beetle wandering around the house, this may be an application.'