Can't compute more than one scenario

Hello,

I am having trouble during the evaluation of docker images on the server. In my docker template, I have a file planner/__ main __.py (see picture attached), which calls my motion planner (motion_planner_interactive) on each scenario and saves the solution in solution_dir.
When I run the docker container locally, I can loop over scenarios I have manually inserted in a scenarios folder and save the solution-xml files to the solutions folder.

However, when I upload the link to the docker image on the challenge page, only one scenario is evaluated. I get the output: Solved Scenarios 1 Failed scenarios 0 Unsolved scenarios 229

Any hints of what the cause may be?

Ps: I have very little docker experience

Hi Leon, have you tried using the provided docker template without any modification? There is a built-in dummy planner which just brakes from the initial state. We could correctly evaluate this on our server (for your reference, solves 43 scenarios from the phase 2 scenarios). Maybe you can try this first and see if you also see 43 successful scenarios?

Hi Edmond,

thanks for the hint. I tested that and 43 scenarios were solved, as you suggested.
However, when I run my own planner for 150 timesteps (as specified in the challenge), I receive a result just for a minority of scenarios (<=11). I now get the output: “Solved Scenarios 4 Failed scenarios 7 Unsolved scenarios 226”. I have been looking into this for a long time and can’t figure out why it works locally on my machine (with training scenarios) and not on the server.
Is there any way for me to see what the error that occurs is?

Hi Leon, maybe you can send me your link to docker image in private, and I can test it on my machine locally as well to see if there is an issue. Also, where do you see the mention of 150 time steps?

I saw it by submitting a solution which had less timesteps:

“Error message: The simulation requires 150 states, but the solution only provides1 time steps!”

Thanks, I’ll send the link to you in private.