So, why didn't they add more sensors, do you think?
I think the
important question is why Tesla (or Elon Musk) is so adamant about not putting lidar scanners into their cars.
As far as why not more sensors: Well obviously part of the reason they don't is the additional expense, and I assume that's the primary reason for refusing to put in lidar.
But from a computer programmer's viewpoint, there is a downside to more and more sensors. Each sensor provides a stream of data input, and that input must be monitored and analyzed in real time. The more data which must be processed and analyzed simultaneously, the slower the entire software decision-making process gets. For example, a recent report said in the case where the Uber car hit and killed a pedestrian, the car did detect the pedestrian, just not in time to stop.
One thing self-driving car designers are going to have to experiment with is finding a sweet spot, a "happy medium", between too little sensor input, which would give an incomplete "picture" of the environment and the objects moving within it, and too much sensor input, which would result in the computer equivalent of sensory overload, slowing decision-making down so much that a slow reaction time by the car would become a danger.
But with the current state of affairs, it's pretty silly to even talk about sensors providing a complete picture. As has been pointed out in a recent article about a second Tesla car hitting another fire truck parked on the highway, semi-self-driving cars currently in production are not built to even
try to react to stationary objects in the lane the car is driving in. (From a lot of comments posted to this article, a great number of people are having a hard time believing that!)
For reasonably reliable (that is, not perfect but pretty good) full autonomy -- Level 4 or Level 5 -- a car is going to have have a SLAM* system. Waymo's experimental test fleet of self-driving cars do have a SLAM system, but nobody -- not even Tesla -- is making a production car which even attempts to create a SLAM.
Tesla's Autopilot+AutoSteer qualifies as advanced Level 2 autonomy. They have a long way to go to get to Level 4, and development of a SLAM system will be part of what's necessary to get there. A SLAM would include detecting the presence and placement of stationary obstacles, up to perhaps 100-200 meters from the car, or at least that far in front of the car, perhaps not as far to the sides and/or behind.
*SLAM stands for Simultaneous Localization And Mapping technology, a process whereby a robot or a device can create a 3D map of its surroundings, and orient itself properly within this map in real time.
)