Safe Streets Rebel’s protest comes after automatic vehicles were blamed for incidents including crashing into a bus and running over a dog. City officials in June said…
Safe Streets Rebel’s protest comes after automatic vehicles were blamed for incidents including crashing into a bus and running over a dog. City officials in June said…
DARPA figures out how to safely drive cars using LIDAR. Musk asked for a self driving car. Engineers come back the LIDAR solution. Musk fires them, says if humans can drive with two eyes, then so can computers. Cameras are cheaper than LIDAR. Second group tries it with cameras, can’t get it to work, asked why they can’t use LIDAR. Second group of engineers is fired. Third group comes up with something that ‘kind of works’. People die. Big companies avoid self driving altogether, even though we have a perfect solution with LIDAR, all because Musk wanted to save a buck and can’t get out of the way of his engineers.
I’ve worked on serious projects involving LiDAR. The LiDAR you need at these speeds and with this resolution cost almost as much as an Electric Car - it’s too expensive to reach wide adoption. But video processing with CNNs/RNNs has proven you can build the same level of data with cameras. You don’t even need binocular cameras now - if objects are moving you can generate binocular data by combining IMU data with time-series imagery.
As I understand it, Tesla’s delays aren’t related to image capture (which is where LiDAR could help). They’re related to trying to find universal actions to take against an almost infinite number of possible scenarios (mostly actions by human drivers).