Rainy Weather Confounds Self-Driving Cars in Michigan State Study
Recent Michigan-based research shows that self-driving cars have a long way to go before they can safely handle nasty weather conditions.
Autonomous vehicles have managed to operate effectively (with some notable exceptions) in the sunny Southwest. But they haven’t fared so well in testing at Michigan State University, where the weather tends to be much less cooperative.
The goal of these Michigan State studies was to test self-driving cars’ ability to avoid hitting pedestrians, cyclists, and other vehicles. This ability is closely tied to the performance of a vehicle’s cameras and sensors. Unfortunately, researchers found that these detectors perform quite poorly in all types of bad weather.
Classy Crossover: Learn more about the new 2019 Cadillac XT4
Rain was especially confounding for self-driving cars tested by Michigan State. During rainfall that was 10 percent of the worst-case scenario for intensity, camera-enabled algorithms were unable to properly spot up to 20 percent of the objects they encountered. At 30 percent of the worst-case scenario, those systems missed up to 40 percent of objects.
Self-driving technology is weakened by cold temperatures, too. Lidar sensors pick up more interference and can’t see what’s in front of the car as sharply when temperatures fall below 10 degrees.
Another problem for autonomous vehicles is seasonal change. To a high-resolution mapping system, a lush summer landscape could look much different when it’s stripped of leaves in the winter. That’s a problem, because self-driving cars rely on these systems to guide them accurately.
In case you’re wondering, Michigan State’s testing apparently didn’t cover snow. But it’s hard to imagine a self-driving car navigating snowy weather safely when even moderate rainfall is too hard to handle.
Luxury Leader: Get the details on the 2019 Cadillac Escalade
News Source: The Truth About Cars