search

ntnu-arl / gbplanner_ros

Graph-based Exploration Planner for Subterranean Environments
BSD 3-Clause "New" or "Revised" License
690 stars 153 forks source link

Using RGB-D Camera #5

Closed engcang closed 3 years ago

engcang commented 3 years ago

Hi. Thanks for sharing this great repo.

I think this repo is really similar with MBPlanner from your lab.

So, the same issue is happening here. Using RGB-D Camera with "kCamera" SensorParams in config files, It continuously returns "No positive gain was found"

[ WARN] [1607462735.649574133, 16.540000000]: [PLANNER_ERROR] No positive gain was found. [ WARN] [1607462735.649621397, 16.540000000]: Planner returned an empty path [ INFO] [1607462736.207201124, 17.040000000]: PlannerControlInterface: Running Planner (kAuto) [ INFO] [1607462736.218772441, 17.052000000]: Planning iteration 3 [ INFO] [1607462736.267533763, 17.100000000]: Set bound mode: kExtendedBound [ INFO] [1607462736.268022768, 17.100000000]: Current box contains Occupied voxels.

I think this repo also has to be fixe as this pull request of MBPlanner. As I made a same issue there, and got fixed from the first author of the paper here

Thank you in advance.

engcang commented 3 years ago

Ohhh Sorry I looked into the code carefully and I found that this pull request of MBP was solved thanks to referring this repo.

And by setting bigger Local and LocalSearch made it work with kCamera.

Thank you!

ShreyanshDarshan commented 3 years ago

@engcang Can you share your Local and LocalSearch values that made it work on kCamera? Also, which RGBD camera did you use on your vehicle?

engcang commented 3 years ago

@ShreyanshDarshan
Hi. I used Rotors simulator as this repo originally adopted it, so I just changed the specification of the camera in component_snippets.xacro rather than using other camera. For the Local and LocalSearch, I used the twice of max_range for X and Y axes, and 1.5, 3.0 for Z axis values each.

But I felt that the bigger values make it work better (e.g., [50, 50, 1.5]), and also resulted in the longer computation time.