| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
I am trying to create a 2D map of a hallway using slam_gmapping and I have not been successful in getting an accurate map.
From what I see, the Odometry appears to be drifting and the slam algorithm is not able to close the loop.
I am trying to figure out if the problem is with the settings of the slam_gmapping parameters or the Odometry data coming from the robot. Any leads, suggestions?
The first thing you need to do before try to make a SLAM is to calibrate. Here in ROS Answers there are several topics about that. Since your problem is that the map is not closing, you can run some tests: Try to go forward into a corridor, rotate 180º and go back to the start point while running gmapping. If the robot creates another corridor on the way back, the problem is the calibration. and you should run it several times until get better results for laser and odometry. If you already have a map, try to send a GOAL to the robot while using amcl_demo and watch the laser in the map (the white line). Again, if the laser exceed the wall, i'm almost sure the problem is with the laser and you need to calibrate it.
Asked: 2014-07-08 11:19:46 -0600
Seen: 953 times
Last updated: Jul 08 '14
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
It would help to state your current parameter set, the computation power you have, which sensor you use, and how the environment looks.
Hokuyo URG-04LX. 2.4GHz dual-core Intel Core i5 processor. The environment is a long hallway with an intersecting side hallway. During mapping, the Robot was tele-operated from one end of the hallway and was brought back to the same location. In the resulting map, the side hallway did not match(there were two - one next to another). I have a schematic. But cannot post it due to insufficient karma. This is default parameters in segbot_navigation
Not able to post the parameters for some reason. But you can view it on gmapping.launch (segbot_navigation) package
Define long. Long hallways are a common problem for laser-based SLAM. If there are no features it will rely on odometry.
228ft X 7.5 ft. Does that mean the odometry is not accurate? Will a UTM-30LX give better results?