Skip to main content
added 288 characters in body
Source Link
flawr
  • 43.9k
  • 41
  • 41

EDIT: I just remembered there was Lab Rat Race: an exercise in genetic algorithms where we did something similar. So maybe it would make sense to take the challenge in a slightly different direction.

  • EDIT: I just remembered there was Lab Rat Race: an exercise in genetic algorithms where we did something similar. So maybe it would make sense to take the challenge in a slightly different direction.

  • EDIT2: There is an almost identical challenge in http://janismac.github.io/ControlChallenges/ (go to the menu with the little squares and choose "Vehicle Steering" or "Vehicle Racing"). After playing around with this I think a combination of golfing code and time would be nice.

EDIT: I just remembered there was Lab Rat Race: an exercise in genetic algorithms where we did something similar. So maybe it would make sense to take the challenge in a slightly different direction.

  • EDIT: I just remembered there was Lab Rat Race: an exercise in genetic algorithms where we did something similar. So maybe it would make sense to take the challenge in a slightly different direction.

  • EDIT2: There is an almost identical challenge in http://janismac.github.io/ControlChallenges/ (go to the menu with the little squares and choose "Vehicle Steering" or "Vehicle Racing"). After playing around with this I think a combination of golfing code and time would be nice.

added 458 characters in body
Source Link
flawr
  • 43.9k
  • 41
  • 41

EDIT: I just remembered there was Lab Rat Race: an exercise in genetic algorithms where we did something similar. So maybe it would make sense to take the challenge in a slightly different direction.


I'd obviously have to write a controller and come up with a good map that is doable with the velocity/direction constraints. Ideally with some graphical output for the participants to debug their car.

  1. Shall there be multiple maps?
  2. Are the discrete actions a good idea? (And are the ranges any good?) I think for the velocity it is nice so the car can come to a full stop in the target area.
  3. Should the steering be limited when the velocity is higher?
  4. Currently it feels like this will end up as just as a compression challenge i.e. who can hardcode a good sequence of steering inputs. I'm not sure this is all that fun. Are there other challenge modes or variations for this setting that would be more interesting? (The other thing I had in mind was giving the players a fixed code-budget and measuring how far they can get in the map.)
  5. Should the cars be allowed to have some memory between each step? (e.g. limited to a certain number of bits) - after the update from above I think it would make sense to allow some (or maybe even an unlimited) amount of memory between each step. But still it would be nice to have some sort of restriction.

I'd obviously have to write a controller and come up with a good map that is doable with the velocity/direction constraints. Ideally with some graphical output for the participants to debug their car.

  1. Shall there be multiple maps?
  2. Are the discrete actions a good idea? (And are the ranges any good?) I think for the velocity it is nice so the car can come to a full stop in the target area.
  3. Should the steering be limited when the velocity is higher?
  4. Currently it feels like this will end up as just as a compression challenge i.e. who can hardcode a good sequence of steering inputs. I'm not sure this is all that fun. Are there other challenge modes or variations for this setting that would be more interesting? (The other thing I had in mind was giving the players a fixed code-budget and measuring how far they can get in the map.)
  5. Should the cars be allowed to have some memory between each step? (e.g. limited to a certain number of bits)

EDIT: I just remembered there was Lab Rat Race: an exercise in genetic algorithms where we did something similar. So maybe it would make sense to take the challenge in a slightly different direction.


I'd obviously have to write a controller and come up with a good map that is doable with the velocity/direction constraints. Ideally with some graphical output for the participants to debug their car.

  1. Shall there be multiple maps?
  2. Are the discrete actions a good idea? (And are the ranges any good?) I think for the velocity it is nice so the car can come to a full stop in the target area.
  3. Should the steering be limited when the velocity is higher?
  4. Currently it feels like this will end up as just as a compression challenge i.e. who can hardcode a good sequence of steering inputs. I'm not sure this is all that fun. Are there other challenge modes or variations for this setting that would be more interesting? (The other thing I had in mind was giving the players a fixed code-budget and measuring how far they can get in the map.)
  5. Should the cars be allowed to have some memory between each step? (e.g. limited to a certain number of bits) - after the update from above I think it would make sense to allow some (or maybe even an unlimited) amount of memory between each step. But still it would be nice to have some sort of restriction.
added 60 characters in body
Source Link
flawr
  • 43.9k
  • 41
  • 41

If the car crashes into a wall, it will be disqualified.

the car with lidar

the car with lidar

If the car crashes into a wall, it will be disqualified.

the car with lidar

Source Link
flawr
  • 43.9k
  • 41
  • 41
Loading