FSM: Correspondenceless scan-matching of panoramic 2D range scans

This repository provides the C++ single-header source file of FSM (include/fsm.h), and testing code.

For using FSM with ROS in order to produce lidar odometry visit fsm-lo.

Why use FSM

Dependencies

CGAL 4.7 FFTW3 boost/random

Building

As always

mkdir build
cd build
cmake ..
make

Run tests located in the dataset directory with

./sm_node A B C D E F G H 0 0 K L M N O P Q R S

e.g.

./sm_node 2 1 0 778 0.2 0.786 0.1 0.0 0.0 0.0 360 360 FSM 200 0 3 10 0 0

where

• A: The number of iterations for the translational component (the larger the location displacement between scans the higher this value needs to be)
• B: How many times to iterate over all instances of dataset
• C: The start sample id; typically 0
• D: The end sample id; here |dataset| = 778
• E: The maximum location displacement. Sample locations are generated uniformly in [-E,+E]
• F: The maximum orientation displacement. Sample orientations are generated uniformly in [-F,+F]
• G: The standard deviation of measurement noise corrupting (real) scans
• H: The standard deviation of measurement noise corrupting virtual scans. Suitable in the scan--to--map-scan matching context; irrelevant in scan-matching
• K: The size of scans
• L: The size of the map. Relevant for scan--to--map-scan-matching tests; irrelevant in scan-matching
• M: Identifier of the method used. Only FSM supported in this repository
• N: The maximum number of iterations over one sampling degree
• O: The minimum sampling degree
• P: The maximum sampling degree. Larger values result in extra accuracy and extra execution time
• Q: The maximum number of recoveries. Larger values result in extra accuracy and extra execution time
• R: Enforcement of a terminal constraint; attempts a recovery if not fulfilled, if true. Set to false for scan-matching
• S: Enforcement of an early gear-up feature, see code for details. Set to false for scan-matching; true for scan--to--map-scan matching

System at a glance