generate synthetic dataset from optimal trajectories

[Armandpl]

alright so we got trajectories and we can import them into blender. here is what's left Augment the data in sim:

• [x] augment data: between each trajectory:
  • [x] randomize the lighting intensity
  • [x] randomize the line colors?
  • [x] swap between a few floor textures? or the moquette texture and random colors?
  • [ ] swap the wall for random colors, or maybe noise?
  • [ ] add noise to camera offset/rotation
• [ ] add ceiling
• [x] match camera and sim
  • [x] choose camera position for the car, measure it
  • [ ] match fov and sensor size. set this from the bpy script
  • [ ] fix the camera profile
  • revert to original one, see if ti fixes the black line
  • then find a new one to remove the purple tint. commit it somewhere?
  • [ ] look into rolling shutter and motion blur. compare sim and recorded data

[Armandpl]

• add gaussian noise to the trajectories
• render all trajectories in one go then split the sequences based on the trajectories

[Armandpl]

• match real cam to blender cam (sensor size? why is the cam in the ground)
  • shutter speed?
• enlever les reflet du sol (et p-e des lignes)
• finir les murs ajouter un plafond
• add random roll to the camera?

[Armandpl]

• track line should be flat

[Armandpl]

temporary solution waiting for the optimal trajectories:

• randomly spawn the car on the track (random x,y in outer bounds) + random yaw
• tune a pid to follow the center line
• make the speed constant
  • add param to the env: constant_speed, adjust action space accordingly
• add a time limit
• add noise in the actions

output will be a list of position and orientations alongside with steering

[Armandpl]

• add noise to camera offset/rotation
• make the line beige
• randomize lighting in sim