About Wall Interaction

[godisreal]

A problem in current simulation is that an agent may go through the wall if he walk towards the wall too fast. In other words, if an agent's moving speed is sufficiently large, the wall repulsion cannot stop him effectively before he touches the wall physically. Thus, he might go through the wall, and this scenario is obviously unrealistic.
From the perspective of pedestrian modeling, it implies that the wall repulsion defined in the traditional social force model does not work out for this scenario.
Any comments are appreciated.

[chraibi]

This is not a problem of your code, but it is a conceptual problem in the social force model.
A quick&dirty fix would be to change the parameters and make the repulsion force stronger.

But again, this is not solving the problem once for good.

[godisreal]

OK. The social force model treats things more like particles than humans. However, I keep using the term of social force because it is a good term for human interaction. Yes. Changing the parameters (e.g.,A and B in Helbing's paper) gives a solution, but not a very good one. The original parameter A=2000 and B=0.08 generate short-range force. Only when agents get very clost to the wall, the force becomes effective. So is there any other good solution you would suggest?

In fact I hope to model the agent more like human. The agent predicts the possible obstacle on the way ahead and thus adjust the desired velocity. I think that will be a better way. Any comments?

[chraibi]

So is there any other good solution you would suggest?

what about you try something much more simpler and more effective? Like a velocity-based model.

[godisreal]

OK. Is there any paper introducing velocity-based model? I know there is speed-dependent social force in FDS+EVAC, but not sure about velocity-based model. I remember you have a paper about "Forced-Based Model" actually.
The speed-dependent force is realized in FDS+EVAC as below. However, I do not quite get the idea. Do you know what this force is used for? It seems that parameter A is adjusted based on HR%SPEED

      ! =======================================================
      ! Speed dependent social force
      ! =======================================================
      IF (C_HAWK >= 0.0_EB .AND. HR%I_DoorAlgo==1) THEN
         HR_A = HR%DoseCrit1
      END IF
      HR_A =  HR_A*MAX(0.5_EB,(SQRT(HR%U**2+HR%V**2)/HR%SPEED))
      A_WALL = MIN(A_WALL, FAC_A_WALL*HR_A)

[godisreal]

Wall interaction is to be upgraded, and we will model agents more like human.
Maybe we can try to read in FDS+EVAC input file (&OBST &HOLE &DOOR &EXIT) to create simple geometry of compartment.

[chraibi]

I mean with Velocity-based model a model that is of first-order not second-order (force-based)

For example this paper
with this implementation

DISCLAIMER: I'm one of the authors 😄