vectr-ucla/direct_lidar_inertial_odometry

keyframe-based loop closure in paper

michael-sz opened this issue · 3 comments

great job! thanks for sharing!

  1. when i read the paper, i found the part of loop closure, but not found during the code . will it be released in future?
  2. if add gps positon as prior factor, the way of lio-sam based on gtsam will be better?

Ah, unfortunately that section is from a follow-up paper (DLIOM) and not from the DLIO paper. We currently have no plans to release that algorithm as of right now. If you do happen to improve on the current DLIO code with a factor graph, I would imagine GPS can only help (assuming it's accurate and properly coded). Let me know how it goes.

got it, thanks for reply.
And if as you convenience , please show me the main distinction between dlio and dilom ,thanks.

tl;dr --

  • DLIO: LIO with fast and accurate motion correction
  • DLIOM: SLAM with innovations in keyframing (resistances to slippage), submapping (looks at overlap / Jaccard index), and mapping (connectivity between keyframes) to increase overall reliability