QUT-Motorsport/QUTMS_Driverless

Perception fusion investigation

Closed this issue · 2 comments

TODO:

  • Establish a suitable method for getting accurate cone detections
  • Implement
  • Optimise
  • Test with optimised LiDAR

Still needed:

  • Option for vision only take over for redundancy.
  • Transforms that change with the model urdf, not hard coded.
  • Make covariances mean something useful / interpreted by SLAM.
  • Nav to use the unknown positions but maybe estimate what colour they could be given their side of the track?

Continued in #178