ethz-asl / voxblox

A library for flexible voxel-based mapping, mainly focusing on truncated and Euclidean signed distance fields.
BSD 3-Clause "New" or "Revised" License
1.33k stars 358 forks source link

Freespace PointCloud Clearing Time #235

Open rghl3 opened 5 years ago

rghl3 commented 5 years ago

Hello, As suggested in #156 and #154 , I implemented the Freespace pointcloud (containing points that are above 3m depth ) Test environment

Issue

Related params

So any idea on why the voxels get cleared slowly when using rays from freespace points ? Thank You.

helenol commented 5 years ago

Hmm... Haven't really experienced this issue before. Could you try playing with the weighting a bit and see if that affects it? use_const_weight to true in the params. Maybe try also with different integrators and see how the issue chnages? i.e., fast since that's the one we usually use. Otherwise could you upload a sample bag with both the real + cleared pointcloud to try out that shows the effect?

rghl3 commented 5 years ago

Hello Helen, Thanks a lot for the response. Tried with fast integrator and the thing is it seems slower than merged. Also, the clearing issue persists in fast mode too. However in merged mode, setting _use_constweight to true did not speeden it up. Once I get back to work, I will upload the bag file. Sorry for that :(

Some more params for reference

Edit 1: Here is the google drive link for the bag file (266 mb)

Topics contained

rghl3 commented 5 years ago

Hi @helenol , Any updates on this?

rghl3 commented 5 years ago

Hi, Regarding the above issue, the rays from freespace points/ points beyond max_ray_length seem to get cleared when considering all points for clearing, by commenting out this line. But the voxels aren't being constructed for sparse obstacle points within the range. Is there any workaround? Thank You. Regards, Raghul.

OPyshkin commented 2 years ago

Hello. Can you please explain how you generate freespace pointcloud?