One of the biggest advantages self-driving cars should be able to give us is a reduction in the number of collisions. For collision detection, self-driving cars use technology such as LiDAR, which fires pulses of light and measures how quickly they rebound, compounding that information with many other environmental variables, or optical camera-based systems.
The technology has had a lot of practice detecting other vehicles, and pedestrians. It has also become pretty good at detecting animals that frequently cross roadways, like deer, elk and caribou, all of which have a fairly standard stride. However, Volvo has indicated there’s one animal that still confuses their autonomous vehicles - kangaroos.
Australia's National Roads and Motorists' Association states that 80% of animal collisions in the country involve kangaroos - more than 16,000 strikes each year, with millions of dollars in insurance claims. The image at right shows just a few vehicles that collided with kangaroos and ended up in Pickles’ salvage inventory.
In this BBC article, Volvo’s technical manager for Australia went over some of the kangaroo-related issues. The problem seems to be one of perception: when the kangaroo is in the air, it appears to be further away than it actually is when it lands. To solve the problem, they’ve engaged in additional data collection where collisions are frequent, and are working on a solution that involves a combination of radar and cameras. The end result should work out well for drivers and kangaroos: Volvo has an ambitious target that no-one is killed or seriously injured in a new Volvo by 2020.