Learn SLAM: Structure of LOAM (0/8)

1. LOAM Structure

P^: the latest lidar scan
Pk : registered point cloud of the Kth lidar sweep

Figure. Diagram of LOAM[1]

2. Key Approaches

(1) Odometry Estimation

In order to get lidar odometry, we need to use methods like ICP. Instead of usign the entire point cloud, LOAM used a smarter way by selecting feature points.

(2) Feature Point Extraction

Equation. Criteria of feature/surface points [1]

Figure
if c>threshold:
sharp edges
else if c < threshold:
planar surface patches

Conditions:
a) The number of selected edge points or planar points cannot exceed the maximum of the subregion
b) None of its surrounding point is already selected.
c) It cannot be on a surface patch that is roughly parallel to the laser beam, or on boundary of an occluded region.

(2) Feature Point Correspondence

Figure. Finding corresepondences [1]

(3) Calculating Residuals

(1) Point to Line Distance

Equation. Point-Line Distance[1]

(2) Point to Surface Distance

Equation. Point-Surface Distance[1]

(4) Motion Estimation

Non-linear Optimization of residuals

Equation. LM Optimization Method [1]

(5) Lidar Mapping

Figure. Lidar Mapping [1]

Q^(k+1): unregistered point cloud (before using LM method for nonlinear optimisation)
Q k+1: registered point cloud

Figure. Lidar Odom Integration[1]

Reference:

[1] Ji Z., Sanjiv S., LOAM: Lidar Odometry and Mapping in Real-time.