search

vipra-uwf / vipra

1 stars 1 forks source link

Pedestrian Dynamics Model #71

Open nrayon opened 1 year ago

nrayon commented 1 year ago

Overview

Started working on getting the Calm Pedestrian Dynamics Model working as intended.

The Pedestrian Dynamics Model should act as a "default movement" for pedestrians when they are not being affected by the human behavior model or the policy model.

Implementation

Implementation of the Pedestrian Dynamics Model includes:

  1. Finding the nearest passenger for each passenger.
  2. Computing propulsion and repulsion of pedestrians.
  3. Calculating velocities of pedestrians.
  4. Checking for deadlocks in Pedestrian Dynamics-related movement.

Current Ideas:

  1. For deadlocks, allow the user to determine whether the simulation should run with deadlocks on or not.

Current Issues

As of this moment, I have no issues accomplishing this task.

rpahle commented 1 year ago

Distance matrix completed Will it be handled anywhere else?

Calm is based on nearest obstacle in direction of motion It is not required to calculate the whole matrix

Nearest Neighbor requires distance matrix. Potential range tree (but will not implement now)

rpahle commented 1 year ago

Working on test for distance matrix

rpahle commented 1 year ago
rpahle commented 1 year ago

Question for @ashok0srinivasan :

rpahle commented 1 year ago

Need values for distance matrinx

rpahle commented 1 year ago

finish tests for distance matrix

ashok0srinivasan commented 1 year ago

From what I remember, the FSU code used a random choice to determine the winner for f the race.

On Mon, Oct 24, 2022 at 4:37 PM rpahle @.***> wrote:

Question for @ashok0srinivasan https://github.com/ashok0srinivasan :

  • race condition to access isle - add randomization? time or acceleration?

— Reply to this email directly, view it on GitHub https://github.com/vipra-uwf/vipra/issues/71#issuecomment-1289649621, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACPNH5BLY6K3EYJQQTUQHVDWE36R5ANCNFSM6AAAAAARFVORDM . You are receiving this because you were mentioned.Message ID: @.***>

--

  • Ashok Srinivasan
rpahle commented 1 year ago
rpahle commented 1 year ago
rpahle commented 1 year ago

Only single isle planes

rpahle commented 1 year ago

blocked by #89

rpahle commented 1 year ago

87 is complete and we can now test with data.

rpahle commented 1 year ago
rpahle commented 1 year ago
GoodenoughR commented 1 year ago

Fixed issue of pedestrian state not being updated properly.

changed calculatePropulsion to use a unit vector for direction to goal. Previous version caused pedestrians to stand still in the aisle. Will more than likely have to change this.

GoodenoughR commented 1 year ago

output

Fixed the pedestrians grouping together at the front of the plane, however there are still some issues.

  1. Commented out the NeighborSpatialTest, to more easily fix issues elsewhere
  2. The pedestrians seem to march in perfect step
  3. The pedestrians look like they are standing too close to each other (though this is probably because the spatial test is commented out)
  4. There is one poor soul stuck in the fuselage in the top left (not sure why he goes up at the beginning)
  5. Peds from the top seem to end up on the bottom of the conga-line in the main aisle and vice-versa