ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | Q&A answers.ros.org

# Laser scan matcher with few scan points

This post is a wiki. Anyone with karma >75 is welcome to improve it.

Hello ROS-Community

I have a question regarding localization with a laser scanner. I am planing to use a Hokuyo Laser Scanner to localize my robot. I tried the laser scan matcher package in a small room and it worked perfectly. The robot is meant to operate on a table 2x3 meters in size. On each corner there are 8x8x8 cm beacons and a large semicircle on the side which must be used for localization. The sensor will see something like the figure below.

   _                       ___________                          _
|_|                     |           |                        |_|
\_________/

_                                                            _
|_|                                                          |_|


Is it possible to use the laser scan matcher package on an environment with few objects to match such as the beacons above? Odometry information will be available. At the moment, the laser is not available, so I can't test it. My guess is, that the laser scan matcher package works mainly with straight lines.

Steve

Appended

Ok, thank you. I will try to limit the FOV of the laser. I thought of tracking the beacons. Since I know their initial position and the laser operates with 10 Hz I can create for each beacon a ROI and ignore the remainig scans. Then it shouldn't (theoretically) matter if there are moving objects around the table. As soon as the laser is available, I will try this. Thank you very much!

edit retag close merge delete

Sort by » oldest newest most voted

This post is a wiki. Anyone with karma >75 is welcome to improve it.

It should work in principle. Is there nothing around it? In a normal rooms you'd usually find "stuff" within 5.6m, so the scan matcher should have more to align on.

Regarding stability: What is your desired application? Because the laser_scan_matcher package does not provide localization, just incrementally matched lasers, i.e. it will drift.

People moving around the table could be really bad as the scanmatcher can't know what to consider and what not. So it will come up with the best match, that might focus on something else than the small table legs. You might need to do something more advanced like actual localization or specific to the task. Maybe you could get away with limiting the FOV of the laser, but you'll definitely have to test that scenario.

more

This post is a wiki. Anyone with karma >75 is welcome to improve it.

The table will be located in a large hall (we are participating in a mobile robot competition and are fairly new to mobile robotics, that is this is our first non virtual robot) There will be about 6 people slightly moving around the table (referee, team members and commentator) but not entering the tablearea. There will certainly be some stationary objects, but I don't know what and in which distance to the table. The primary goal is to know where the robot is and to navigate it to several predefined points on the table with a certain orientation. We know the inital pose of the robot. I noticed this drift in the first test but hoped it would not affect the application. So maybe the approach with the laser_scan_matcher isn't the best choice. I would like to not completely rely on the odometry because we are using omniwheels and should navigate to the positions as precisely as possible. But until now I couldn't think of a better solution to acomplish this task.

more

( 2013-02-14 22:27:35 -0500 )edit