PostCrashCalculation delivers wrong values

In the attached example simulated with openPASS v0.10, the PostCrashCalculation computes wrong values in the simulationOutput. Indeed, the logged parameters are all equal to zero, as shown in the screenshot. This is definitely buggy, since there has been a collision with a delta velocity equal to 26 km/h.

Configs_openPASS0.10_PostCrashBug.zip

added Bug label

changed the description

The PostCrash calculation in collisionDetection_Impact_implementation seems not to work for pedestrians according to your definition, because it requires vehicle parameters or even input from collision specific vehicle components, as far as I know.

We did some PCM simulations with pedestrian cases with pedestrians as PCM agents using the static systemConfig Agent_Copy.xml.

The PostCrash calculation works here (for 0.8, 0.9), but I need to test this for v0.10

added Collision Detection label

Thanks for your reply! This explains it all. But again we need a documentation I still find the output with zeros quite misleading...Maybe the PostCrashCalculation should be deactivated, if no calculation is possible, e.g. with Pedestrians.

I agree: we need to further consolidate end-to-end for which cases the PostCrashCalculation works (input type, agent type..) - and then document this properly.

But first, let's consolidate the requirements and concepts how we handle Collision Detection within the new Mantle architecture.

assigned to @jdobberstein

set weight to 1

changed milestone to %Release v1.0.0

@mhammouda Is there any particularity about the cases you are running? I have not conduct a thorough investigation of the topic, but in my experience the PostCrashCalculation works on and off. In my case, I have seen this issue in pedestrian impacts where the impact happens close to the vehicle fender. I am also using v0.10

@dperezrapela yes, for the moment the PostCrashCalculation cannot handle collisions with pedestrians correctly. Please refer to @jdobberstein's comment here https://gitlab.eclipse.org/eclipse/simopenpass/opSimulation/-/issues/183#note_1062113

Thank you @mhammouda, my only comment is that the error does not occur in all pedestrian impact but only in some. In my case, it seems to fail when the pedestrian hits close to the fender of the vehicle.

@dperezrapela Alright! No, I honestly did not investigate this further. But this may be the case... Maybe @jdobberstein can consider your input for further investigations.

Thanks for your feedback, we will definitely consider this in the re-factoring.

Running the E2E-Test Pedestrian_Trajectory lead to meaningful simulationOutput (see below), so your overall setup is fine, but these collisions close to the fender seem to be edge cases where the interpretation of the contact plane for PostCrash_Calculation is not possible.

The impact model behind these calculations was developed and tested with full-overlap vehicle-to-vehicle collisions, so we need to understand and describe the boundaries better, see this paper about the concept: Kolk et. al.

mentioned in merge request !159 (merged)

covered by updated documentation (!159 (merged))

closed