The drive to develop automated vehicles, “AV” for short, has gone into high gear this past year. From the US to Europe to Asia, AV technology is advancing by leaps and bounds. But in one place the road to automation has taken an interesting turn.
A huge factor in the production of safe automatic vehicles involves the development of reliable recognition technology. Vehicles need to be able to distinguish which objects in front of them are hazards (like pedestrians) and which are not (think tumbleweed). A major function of recognition technology, then, is to focus on the shape and movement of various potential hazards to determine what they are – and that has brought advancements in AVs recognizing animals like deer, elk, and moose. (We’re not sure about rabbits and chipmunks.)
But in Australia engineers seem to be stumped by one of their own: the kangaroo.
Our friends over at Marketplace explain that “Kangaroos are much trickier – faster and more shape-shifting – than other animals that wander onto the road.” Deer and moose are big and generally move slowly, adds University of Washington computer scientist Pedro Domingos, making them easy for a vehicle’s vision system to recognize them as such.
Kangaroos, on the other hand, jump across the road much more quickly, keeping low and essentially changing shape as they hop, which makes it harder for the vehicle’s computer vision system to understand that “Yes, this is a kangaroo and not a wind-blown plastic bag.”
Of all animal-related accidents in Australia, eighty percent involve kangaroos. Kangaroo-recognition, then, is a critical aspect of automated vehicle technology.
That’s the challenge in Australia, anyway. Here in the States, our engineers are scratching their heads wondering how to deal with all the direction-changing squirrels in the road.
Maybe our engineering issues will be easier to overcome.