Agents get stuck against walls and remain immobile

[thombie]

In the following simulation setup, some agents eventually become stuck after being pushed against a wall. Even when there is open space around them, they remain frozen in place and never resume movement.

]

video.MOV

Simulation settings:

dt = 0.05[Same problem occurred when dt=0.01

Agent settings:

Model: AnticipationVelocityModel

All other parameters are set to their default values.

The agents are generated automatically at a fixed point whenever surrounding space becomes available.

Is there any known way to prevent this issue?

The attached image shows the positions and radii of the waypoints, and the video captures the trajectories of the agents during the simulation.

[chraibi]

I suspect a bug in the model. Can you run the same setup but with the CollisionFreeSpeedModel?

Can you upload the sqlite file?

[thombie]

Thank you for the quick response.

I tried switching to CollisionFreeSpeedModel, and it seems to have resolved the issue.

Here is the .sqlite file (I had to rerun the simulation since the original data was lost) and a video showing the agent trajectories during that simulation.[The files were too large and could not be uploaded to GitHub]
link: https://11.gigafile.nu/0918-f5a3e432fdcb05f33fdbbb1ec781b7c2
[Please click "まとめてダウンロード" in the link]

[chraibi]

Thanks for the files. I'll take a look. AVM is relatively untested yet. I might have an idea why this happens.
Do you mind sharing the coordinates of the geometry?

As a side note, you can use jupedsim-visualizer?

[thombie]

default_geometry_coordinates = [
    [(19.35, 56.24),(24.38, 64.50),(49.15, 49.15),(41.80, 42.57)],
    [(41.80, 42.57),(43.86, 31.48),(48.63, 32.51),(49.15, 49.15)],
    [(48.63, 32.51),(49.15, 49.15),(50.31, 50.57),(58.31, 46.05)],
    [(48.63, 32.51),(51.21, 20.00),(67.08, 40.25),(58.31, 46.05)],
    [(51.21, 20.00),(56.63, 17.54),(69.40, 38.83),(67.08, 40.25)],
    [(58.31, 46.05),(67.08, 40.25),(88.03, 77.40),(72.11, 69.14)],
    [(72.11, 69.14),(73.01, 69.84), (79.78, 81.86),(79.08, 82.56),(85.14, 79.85),(88.03, 77.40)],
    [(79.08, 82.56),(85.14, 79.85),(89.78, 86.95),(85.40, 92.49)],
    [(79.85, 95.33),(79.85, 103.97),(87.98, 102.56),(85.40, 92.49)],
    [(102.81, 70.82),(106.04, 72.24),(111.71, 68.11),(110.04, 65.40)],
    [(89.78, 86.95),(92.88, 90.56),(106.04, 72.24),(102.81, 70.82)],
    [(85.40, 92.49),(92.75, 93.91),(92.88, 90.56),(89.78, 86.95)],
    [(85.40, 92.49),(87.98, 102.56),(97.78, 100.62),(92.75, 93.91)],
    [(87.98, 102.56),(97.78, 100.62),(114.42, 103.59),(114.04, 106.17)],
    [(114.42, 103.59),(97.78, 100.62),(92.75, 93.91),(92.88, 90.56),(106.04, 72.24),(111.71, 68.11)],
]

are the coordinates of walkable area(the walkable area is union of all polygons generated from each list of coordinates)
goal_1 = simulation.add_exit_stage([
(113.792, 106.106),
(113.502, 106.082),
(113.022, 103.660),
(114.308, 103.700)
])
goal_2 = simulation.add_exit_stage([
(109.723,69.131), (110.907,68.319),(109.883, 65.822), (108.334, 67.112),
])

are the coordinates of the exits.

[thombie]

In my environment, jupedsim-visualizer does not finish generating animations, so I am using matplotlib's simple animation instead.[Thank you for informing]

[chraibi]

Can you upload the sqlite file here?
https://fz-juelich.sciebo.de/s/FNfvCsEpMYJf3bL

Your file can't be downloaded (it takes more than 20 hours and is instable) :-(

[thombie]

Thank you for letting me know where to upload the file.

[thombie]

Should I upload the sqlite file with every_nth_rate=10or100 or trajectories after the agents stuck?[the file is extremely large: about 21GB and sciebo says it's too large]

[chraibi]

every_nth_rate=100 should be enough

[thombie]

Thank you. I successfully uploaded the sqlite file to the link.

[chraibi]

Thanks for the coordinates. I could reproduce the bug.
I'm not sure yet what it is, but I suspect the wall handling is buggy around sharp corners.
I'll try to fix it. Meanwhile thanks for your help!

Screen.Recording.2025-06-11.at.19.54.47.mov