search

RobotLocomotion / gcs-science-robotics

Motion Planning around Obstacles with Convex Optimization by Marcucci et al, 2023
BSD 3-Clause "New" or "Revised" License
150 stars 19 forks source link

Running time of the planner #7

Closed caseypen closed 5 days ago

caseypen commented 7 months ago

Hi, I tried the function of plan_through_buildings in the /reproduction/uav/helps.py and examples/uav_planning.ipynb. I used a Desktop with a pretty decent CPU(i7-13700F), but the running time of the planning is very long (over an hour!!). Is this right? Can you provide some "lighter" examples for the user to just understand the whole picture well? I thought the whole method was in online fashion as described in the paper, right？ It will be better for the example to integrate the whole planning process including generating the convex feasible space with IRIS (I think, right?). I know the optimization part is workable in the example "reproduction/min_length_vs_min_time.ipynb". So does generating convex space takes so long time?

Thanks !

RussTedrake commented 5 days ago

Sorry that we didn't respond sooner. The running time should be very fast; it seems that something is wrong.

I have an (improved) reproduction of that notebook available here: https://deepnote.com/workspace/Underactuated-2ed1518a-973b-4145-bd62-1768b49956a8/project/10-Trajectory-Optimization-05031f8c-3586-4b47-be79-7f4893cf2f9d/notebook/gcs_quadrotor-dcbe8a11149940f8bfd9dcbb8b93df0e

Even on deepnote, using open-source convex optimization tools, it solves in about 1 second.