search

strands-project / strands_movebase

A repository for all the STRANDS-augmented movebase, including 3D obstacle avoidance, etc.
MIT License
10 stars 20 forks source link

[strands_movebase] Don't add obstacles at edge of laser's FOV #26

Closed nilsbore closed 10 years ago

nilsbore commented 10 years ago

This pull request adds a new node, remove_edges_laser that removes a small angle at the beginning and end of the laser scan. This new scan is then used to add obstacles in the costmap while the old, full, scan is used for clearing them. This should solve https://github.com/strands-project/strands_movebase/issues/25 .

Since I am unable to try this out on Rosie atm (going on 48 hours now=P) and we are performing a marathon, I would encourage those that think that this affects the performance of their robot's navigation to try it before we merge. @BFALacerda @lucasb-eyer

lucasb-eyer commented 10 years ago

Will also not try it before the marathon, but hopefully won't forget about it thereafter!

bfalacerda commented 10 years ago

Bob is using it

nilsbore commented 10 years ago

Bob: King. If there are still lingering obstacles added by the laser, you can increase the laser_angle_cutoff parameter.

nilsbore commented 10 years ago

@BFALacerda We've tested this on Rosie now also and I can't get obstacles to linger though it wasn't exactly consistent before either. Any objections to merging this?

bfalacerda commented 10 years ago

we can merge it, it doesn't hurt. I'm also not 100% sure that it avoids the addition of lingering obstacles, but we had less problems since we added this.

We're also back to using the costmap reloading recovery behaviour so it's hard to know what improved the performance. Anyway, merging now and will do further testing later

lucasb-eyer commented 10 years ago

I'm also not 100% sure that it avoids the addition of lingering obstacles

I think the current costmap approach of ROS is fundamentally flawed and we'll never completely get rid of lingering points. But this sounds like it will (highly?) reduce their probability.