About mapping

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

About mapping

plaid
Hello. Sometimes it goes wrong as shown in the picture below. It is very strange. Why is this? This is not the case, it happens even in normal situations.
Reply | Threaded
Open this post in threaded view
|

Re: About mapping

matlabbe
Administrator
Hi,

Is it icp_odometry? It looks like it is a TOF camera -like point cloud and very flat surfaces in the environment (is it a simulated environment?). If ICP doesn't have enough geometric constraints in all three axes (x-y-z), it will drift fast.

To be more robust to that kind of environment, a larger field of view sensor/camera (or multiple cameras) would be required.

cheers,
Mathieu
Reply | Threaded
Open this post in threaded view
|

Re: About mapping

plaid
No, I don't. This is wheel odometry Yes, we ran the simulation in a flat environment, but that's not icp odometry Is there still a possibility of drift in a flat environment?

And let me ask you one more question. Is it possible to access the information on octmap? I would like to use this to make a successful decision. Thanks in advance.
Reply | Threaded
Open this post in threaded view
|

Re: About mapping

matlabbe
Administrator
If it is wheel odometry alone, it is odometry drifting. If the map is okay until a loop closure is found, then destroyed like this, you may adjust your odometry covariance and set it lower (I would suspect your odometry uses Identity matrix if so). This will avoid shewing the map like this.

Rtabmap publishes an OctoMap.