This (2020) is more uptodate than 2021 one.

<aside> đŸ’¡ Author: Udit Singh Parihar

</aside>

  1. Fig. a is the Ground truth trajectory. The robot started at coordinates (0, 0) and was steered on a squared path, revisiting the first part of the trajectory.
  2. Fig. b is the map generated from odometry measurements only. The robot positions are depicted in blue. Green links between two poses are loop closure constraints requested by the place recognition system.

SLAM Representation

  1. Simultaneous Localization and Mapping (SLAM) representation as Conditional Dependency Graph.
  2. White nodes represent unobservable hidden variables (i.e. Robot’s positions and Landmarks).
  3. Dark nodes represent observable random variables (i.e. Sensor Measurements).
  4. For SLAM problem, quantity we want to estimate:

$$ P(X_T, M | U_T, Z_T) $$

Bag of Visual Word (BoVW)

What is Bag of Visual word?