Autonomous vehicles are becoming more common on roads.  These cars use different technologies to see and understand their surroundings.  There are six levels of vehicle autonomy.  These levels range from zero to five.  Level zero means no automation.  Level five means the car can handle any situation without human help.  Tesla's cars are at level two.  This means they can steer and speed up on their own.  However, a driver must always be ready to take control.  Tesla uses cameras and vision processing for features like Autopilot.  These systems help the car see and make safe choices.  But, all self-driving cars need a way to see and understand their environment.  Some companies use LiDAR.  LiDAR is great for making 3D maps in tough conditions.  Tesla prefers vision-based systems.  These systems can identify objects and improve over time using AI.  Plus, they are cheaper.

    A test was conducted with an older version of Tesla's Autopilot.  The goal was to see if a photo of an open road could trick the Tesla's vision-based system.  The result?  The Tesla drove straight into the fake road.  It was like something out of a comedy sketch.

    Before this test, the creator showed how LiDAR cameras work.  He took a LiDAR camera on a Disney ride in near-darkness.  LiDAR can create detailed 3D maps even in complete darkness.  This is why some car companies think LiDAR is crucial.  Tesla has criticized LiDAR systems.  They say LiDAR is "dumb" and can't identify objects.  However, LiDAR can detect objects that blend in with the road.  This is why a LiDAR-equipped car didn't fall for the photo wall trick.  LiDAR also handles situations that vision-based cameras struggle with, like heavy fog.

    It's uncertain if Tesla's newest system would have done better.  Critics have pointed out that vision-based systems can be tricked by visual illusions.  They also struggle in bad weather.  This raises questions about how to safely create a fully autonomous vehicle.  Would it need a mix of different imaging technologies?

    Running into a photo wall is unlikely.  But, driving in unpredictable situations is not.  Level four and five autonomous cars will need advanced camera technology.  This technology must handle everything the real world throws at them.  The road ahead for autonomous vehicles is complex.  It's a mix of exciting possibilities and challenging obstacles.

The Road Trick That Stumped a Tesla

Actions