Mud in your eye: Self-driving cars tackle dirt on sensors
- Joann Muller, author of Axios What's Next
Argo AI's mud cannon flings dirt at the sensors on a self-driving car. Image: Argo AI
Autonomous vehicle developers are deliberately blinding their cars to perfect various sensor-cleaning strategies.
Why it matters: Like humans, self-driving cars must be able to actually see the world around them to make decisions. But rain, dirt, debris and even insects can obscure their vision, requiring systems that can automatically clean the sensors.
- Companies have been testing a variety of cleaning methods, from blowing small puffs of air on cameras and lidar sensors to squirting them with liquids and then wiping them with tiny squeegees.
What's happening: To help with the research, engineers at one company, Argo AI, even jerry-rigged a "mud cannon" to splatter its test cars with a pudding-like substance.
- The goal was to simulate mud splashing from a passing car — without damaging the sensors.
- During six weeks of trial and error, Argo AI constructed a hood-mounted gizmo from PVC pipe, sprinkler valves and a portable air compressor that could fling a carefully crafted mix of water and scratch-proof dust at the car's sensors.
- The end result is like a "simple potato gun, with a sprinkler valve."
What to watch: 3M is developing a replaceable, superhydrophobic film that would act like a waterproof screen protector for lidar systems and other vehicle sensors.
- 3M scientist Jonah Shaver, who developed the film as a side project, says the company is targeting several industries, including transportation, drones, agriculture and mining.
- "We really think of it for use in any place you want a machine to see a little better."
