3 ms·

This is what I'm trying to avoid. The complexity there is enormous. Webcams have rolling shutters and decent latency, pointing a lidar from a camera detection i

This is what I'm trying to avoid. The complexity there is enormous. Webcams have rolling shutters and decent latency, pointing a lidar from a camera detection is not trivial - there's distortion and delays and everything would need to be very fast and steady. It's a very hard actuation and sensing problem. How do I even calibrate it so I know the lidar is pointing at the right pixels? Easier to use a few cameras on my roof (for wide baseline) and do triangulation. Maybe with GPS edge trigger. But the appeal of a small antenna and MCU is also a factor. A 5W transmitter is small and easy to use, there isn't any traffic to deal with just occasional cars, Doppler effects are fairly well understood and radar has been around for 80 years so there's some retro cool too.

solardev 2y ago
Would it be very different from those DIY motion-tracking turrets like https://hackaday.io/project/18665-motion-tracking-airsoft-turret https://hackaday.io/project/18665-motion-tracking-airsoft-tu... or https://courses.ece.cornell.edu/ece5990/ECE5725_Spring2018_Projects/fy57_xz522_AutoTurret/index.html https://courses.ece.cornell.edu/ece5990/ECE5725_Spring2018_P...?
- jvanderbot 2y ago
  Yes it would. Those track targets at a few feet, meaning all that shaking it's doing when it's "Pointing" would just wave the laser around in the general direction of the car that is much further away. from 4-5 feet away, a person with 2 foot width ~= .5 radians or 28 degrees of pointing accuracy needed, and no tracking accuracy needed since it doesn't seem to follow targets well. (that second link the target is all of 2 feet away and 1 foot width). 100m away, an aproaching car has 2m width ~= 1/50 radians, or ~1 degree of pointing accuracy. More to keep on center of mass. And a car at that range would be about 20 pixels wide. That doesn't inspire feelings of robust visual tracking the way "Big blue blob of paper that fills 20% of the view" would.
  - solardev 2y ago
    Makes sense. Thanks for the explanation!

solardev 2y ago

Would it be very different from those DIY motion-tracking turrets like https://hackaday.io/project/18665-motion-tracking-airsoft-turret https://hackaday.io/project/18665-motion-tracking-airsoft-tu... or https://courses.ece.cornell.edu/ece5990/ECE5725_Spring2018_Projects/fy57_xz522_AutoTurret/index.html https://courses.ece.cornell.edu/ece5990/ECE5725_Spring2018_P...?

jvanderbot 2y ago
Yes it would. Those track targets at a few feet, meaning all that shaking it's doing when it's "Pointing" would just wave the laser around in the general direction of the car that is much further away. from 4-5 feet away, a person with 2 foot width ~= .5 radians or 28 degrees of pointing accuracy needed, and no tracking accuracy needed since it doesn't seem to follow targets well. (that second link the target is all of 2 feet away and 1 foot width). 100m away, an aproaching car has 2m width ~= 1/50 radians, or ~1 degree of pointing accuracy. More to keep on center of mass. And a car at that range would be about 20 pixels wide. That doesn't inspire feelings of robust visual tracking the way "Big blue blob of paper that fills 20% of the view" would.
- solardev 2y ago
  Makes sense. Thanks for the explanation!