Robots learn to explore dense terrain from insects

An international research team involving Johannes Kepler University Linz has transferred insects’ visual behaviour to robotic dogs, ORF reports. By imitating insects’ motion parallax, autonomous robots can move through undergrowth more effectively and detect obstacles, improving their ability to explore terrain visually.

Grasshoppers and praying mantises use sideways body movements to judge distances. This motion parallax allows them to quickly detect nearby obstacles and estimate the distance to prey. The approach could, for example, assist rescue operations in dense forests.

Robotics has previously relied on 2D cameras or 3D sensors, which often fail when scenes are obscured.

Insect behaviour meets artificial intelligence

The research team equipped four-legged walking robots with upper-body movements resembling the insects’ peering behaviour. By superimposing individual images, the robots create a virtual lens that blurs foreground vegetation while bringing concealed objects into sharp focus.

Biology and robotics combined

The researchers mathematically modelled the peering behaviour and transferred it to the movement mechanics of walking robots. This imitation enables artificial intelligence models to identify obscured scenes more reliably.

“Insects show us that perception is not a passive process but an active behaviour. By enabling robots to behave like insects, we are solving one of the most persistent problems in visual exploration in open terrain,” said JKU professor Oliver Bimber.