Understanding occupancy grid related parameters and concepts
Posted by Jeonghwan Park on Aug 20, 2024; 4:23am
URL: http://official-rtab-map-forum.206.s1.nabble.com/Understanding-occupancy-grid-related-parameters-and-concepts-tp10387.html
Hello Mathieu, grateful RTABMap user here.
First of all, i want to express my gratitude for open-sourcing such a competent package for everyone's use! My experiences confirm that RTABMap really is powerful in terms of both configurability and mapping/localization performance.
With that in mind, I wish to ask some questions about how to understand some parameters and concepts related to occupancy grids. My robot configuration includes a 2D lidar and RGB-D camera, and my goal is to create 2D occupancy grid map for navigation purposes.
1. About parameters "Grid/DepthDecimation" and "Grid/ScanDecimation". How does the decimation of RGB-D images and laser scans actually work, and what are their potential impacts on mapping/localization performance?
2. From my understanding, parameters such as "Grid/Footprint~", "Grid/MaxGroundAngle", "Grid/MaxGroundHeight" and "Grid/NormalsSegmentation" are valid only when RGB-D and/or 3D lidars participate in grid mapping. Is this correct?
3. When I set "Grid/Sensor" to 2, how are 2-D lidars and RGB-D cameras combined to formulate the grid map? Does the depth data from RGB-D cameras get projected to the ground to draw additional obstacles?
4. How can I best understand the "GridGlobal/ProbHit and ProbMiss" parameters?
5. I guess that "ICP/DownSamplingStep" and "ICP/VoxelSize" parameters are applied to original laser scans to be used for 2D lidar ICP odometry. Is this correct? And how can I best understand "Uniform Sampling" for scans?
6. Are the parameters with prefix "Mem/LaserScan" independent with similarly named parameters from Grid related parameters?
7. Is the parameter "RGBD/CreateOccupancyGrid" parameter unrelated with the "Grid/Sensor" parameter?
8. From my understanding, "Proximity detection" is a scan-matching based map matching technique where past scans are combined and compared with the current scan to establish connections between the current pose and the map. Is this correct?
9. About the "Reg/Strategy" parameter. Is this the switch for how to calculate the loop constraint? If a 2D lidar is available and this parameter is set to 1(ICP), Is the loop constraint calculated using scan matching instead of RGB-D data?
That was a lot of questions
which demonstrates the configurability and versatility of RTABMap, I think! Mighty thanks for the development and support of RTABMap!
URL: http://official-rtab-map-forum.206.s1.nabble.com/Understanding-occupancy-grid-related-parameters-and-concepts-tp10387.html
Hello Mathieu, grateful RTABMap user here.
First of all, i want to express my gratitude for open-sourcing such a competent package for everyone's use! My experiences confirm that RTABMap really is powerful in terms of both configurability and mapping/localization performance.
With that in mind, I wish to ask some questions about how to understand some parameters and concepts related to occupancy grids. My robot configuration includes a 2D lidar and RGB-D camera, and my goal is to create 2D occupancy grid map for navigation purposes.
1. About parameters "Grid/DepthDecimation" and "Grid/ScanDecimation". How does the decimation of RGB-D images and laser scans actually work, and what are their potential impacts on mapping/localization performance?
2. From my understanding, parameters such as "Grid/Footprint~", "Grid/MaxGroundAngle", "Grid/MaxGroundHeight" and "Grid/NormalsSegmentation" are valid only when RGB-D and/or 3D lidars participate in grid mapping. Is this correct?
3. When I set "Grid/Sensor" to 2, how are 2-D lidars and RGB-D cameras combined to formulate the grid map? Does the depth data from RGB-D cameras get projected to the ground to draw additional obstacles?
4. How can I best understand the "GridGlobal/ProbHit and ProbMiss" parameters?
5. I guess that "ICP/DownSamplingStep" and "ICP/VoxelSize" parameters are applied to original laser scans to be used for 2D lidar ICP odometry. Is this correct? And how can I best understand "Uniform Sampling" for scans?
6. Are the parameters with prefix "Mem/LaserScan" independent with similarly named parameters from Grid related parameters?
7. Is the parameter "RGBD/CreateOccupancyGrid" parameter unrelated with the "Grid/Sensor" parameter?
8. From my understanding, "Proximity detection" is a scan-matching based map matching technique where past scans are combined and compared with the current scan to establish connections between the current pose and the map. Is this correct?
9. About the "Reg/Strategy" parameter. Is this the switch for how to calculate the loop constraint? If a 2D lidar is available and this parameter is set to 1(ICP), Is the loop constraint calculated using scan matching instead of RGB-D data?
That was a lot of questions
which demonstrates the configurability and versatility of RTABMap, I think! Mighty thanks for the development and support of RTABMap!
| Free forum by Nabble | Edit this page |