Trouble avoidance

Making locusts watch battle scenes from science-fiction films such as Star Wars may eventually help cars and planes steer themselves out of trouble. Researchers in the United Kingdom and Switzerland have already exploited this bizarre research as part of a plan to build a robot that can avoid approaching objects. They hope the robot's neural network program, which mimics part of the locust's brain, will be the basis of a collision-avoidance system for vehicles. The conventional approach to creating such systems involves using radar or

infrared detectors and requires very heavy-duty computer processing.

Insects manage to dodge objects despite their poor vision and basic brains. In locusts, the key to this ability is a large neuron behind each of their compound eyes called the lobula giant movement detector (LGMD). The LGMD is thought to be partly responsible for triggering the insects' escape jumps and steering responses during flight, said Claire Rind, a neurobiologist at the University of Newcastle, north-east England. So copying the behaviour of the LGMD could help create fast and efficient avoidance systems for vehicles such as cars and planes.

To build up a picture of how this neuron works, Rind showed images of rapidly approaching objects to a locust while monitoring the neuron's activity. She chose scenes from "Star Wars'' because the film has particularly good shots of objects, such as fighter ships, heading for the viewer during dogfight sequences. Working with Mark Blanchard and Paul Verschure of the Institute of Neuroinformatics at the Federal Institute of Technology in Zurich, Rind built a robot with a vision system based on the locust's. It has a camera with a resolution of just 20 by 20 pixels that closely imitates the locust's grainy vision. As in the locust, signals from these makeshift photoreceptors are subjected to some pre-processing by a neural network - providing some basic "edge detection'' - before they reach the robot's movement detector which is designed to respond in the same way as the LGMD.

The researchers found that their robot avoided objects for 91 per cent of the time, although it had only a short time to react. Nigel Clarke, principal engineer at car company Jaguar's radar applications research unit in Coventry, English Midlands, has been keeping a close eye on Rind's progress. Clarke said: "We certainly think there is a great deal of mileage in using techniques that the brain uses to calculate these things, such as detecting moving objects in a scene.'' Blanchard believes that traditional approaches to collision avoidance - such as infrared or laser rangefinders and ultrasonic reflections - are too easily confused by spurious signals.

"Most biological systems can see,'' Blanchard said, "so it's at least one approach that should be exploited in trying to make machines explore for themselves.'' But Rind says more research is needed before the technology can be exploited commercially. She now wants to monitor the locusts' response to natural stimuli, instead of films, and work out how the speed of approaching objects affects behaviour. Elsewhere, robotics researchers at Sussex University, southern England, are examining how honeybees navigate efficiently with very small brains because they hope to use similar systems to allow robots to navigate autonomously.

Rind explains the thoughts behind this: "Insects in general are good sources of inspiration for visual solutions to some challenging situations in machine vision for two reasons. First, insects operate in the same world as more complicated animals, so the challenges from the environment are the same but are met with a smaller number of nerve cells improving our ability to get to grips with what goes on in their nerve cell circuits and exploit it. Second, insect eyes are particularly good at seeing rapid motion rather than static detail, making insects ideal as inspiration for controlling actions in real time for a fast mobile robot or vehicle.''