search

SteveMacenski / slam_toolbox

Slam Toolbox for lifelong mapping and localization in potentially massive maps with ROS
GNU Lesser General Public License v2.1
1.5k stars 497 forks source link

Detecting Smaller Object #705

Closed lesu0330 closed 3 weeks ago

lesu0330 commented 3 weeks ago

Feature description

Create a way so that the slam can map smaller objects. This will be able to solve problems when needing to use slam for smaller obstacles. Currently the slam toolbox does not consider

Implementation considerations

I do not know how this feature can be implemented completely but maybe have additional parameters in the param file or be able to adjust the ceres solver.

SteveMacenski commented 3 weeks ago

What do you mean "smaller objects"? Are these objects that your lidar is not picking up? That would seem more a concern with lidar resolution w.r.t. the obstacles you want to map issue than the algorithm itself. But please clarify :-)

lesu0330 commented 3 weeks ago

I meant for example if there is a box that is for example 10 cm. It considers as a obstacle and maps it. Is there a way to make it so that it does not consider it as an obstacle and don't map it?

SteveMacenski commented 3 weeks ago

No, there is no way to do that with the SLAM problem. The better and more common solution is to have a detector before SLAM which is detecting the small obstacle in your sensor reading and then publishing sensor data with that obstacle removed. Then that topic can be used by SLAM and it never knew about the small obstacle in the first place. These are pretty common to have sensor filtering and processing pipelines for depth/lidar measurements before ever hitting the SLAM or obstacle avoidance algorithms.

I think that's the right/best solution!

lesu0330 commented 3 weeks ago

Just to clarify what you have said. Are you indicating that the only way to detect smaller object or ignoring an object is to change the lidar measurements/performance. Make it worse if I wish to ignore the obstacles and make it better if I do want to detect it? Then may I also ask what the tolerance and trust areas do in the ceres_solver.cpp?

SteveMacenski commented 3 weeks ago

lidar -> pre processing stages -> {slam, perception, obstacle avoidance, etc} is the common pipeline. If you want to ignore some data based on size or detected status, you should have a preprocessing stage for that.

lesu0330 commented 3 weeks ago

So there will need to be a preprocessing stage before SLAM? That is not part of the slam toolbox repository?