|
|
All data analysis and visualization have been performed on a Intel CPU based PC running Ubuntu Linux 8.04. For a preliminary overview, we plotted the displacement of each person from his/her original position with time. Custom code was written in Python for this purpose. The result is shown in Figure 1.
|
Figure 1: Plot of displacement vs. time of individual in the building
It clearly showed that there was a mass movement in the building starting around the time frame 370 or so which is assumed to be the explosion. Having gained this quick insight, a custom Python software package called ESR (Evacuation Scenario Renderer) was developed to generate 2D visualization of each of the time frames in the dataset. This allowed detailed investigation of movements of people in the building. Figure 2 shows a screen-shot for time 374.
|
Figure 2: Image rendered by ESR software for time frame number 374. Click on image to get full resolution version
The open source image viewer called gqview was used to move forward and backward through the set of rendered images to perform detailed analysis of the movements of people with time. We also combined all the ESR rendered 2D visualizations to create an animation which, along with the gqview based analysis, provided valuable information regarding the evacuation scenario. To download and view these animations, click on the following links:
These animation files were tested under Linux using mplayer, xine and vlc software.
While doing the analyses, we realized the need for a tool to for visualizing the motion track of individuals. A custom interactive 3D track viewer called ITView (Individual Track Viewer) was developed using the Python scripting language. It is an interactive 3D track viewer that shows tracks for selected individuals as a 3D stack. Figure 3 shows a screen-shot of this tool.
|
Figure 3: A screen-shot of ITView software tool. Click on the image for a full resolution image
In this figure, the tracks for RFID values 17, 27, 29, 38, 44 and 53 are displayed as lines in different stacked layers with the building layout at the bottom. The tracks cover the entire time span of the provided dataset. A transparent plane can be drawn through any of the stacks for easy visual association of the displayed RFID numbers with the corresponding tracks. The part of a track that has already been traversed at a given time is colored green, and the part that is yet to be traversed is colored red. This allows easy comprehension of the time frames when different individuals started moving as well as the relative distances traveled by individuals by a certain time.
With these tools, we came to the following understanding of the evacuation scenario. Note that all grid cell coordinates are stated as (col, row) tuple.
Most people were more or less stationary until the time-frame 375 when large simultaneous movements started to take place. This indicates that the explosion happened at time frame 374. Three of the early movers were #29, #44 and #56. They start moving before most in the building. #29 and #44 eventually meet near the grid cell (54, 15) at time frame 262. This looks quite suspicious, specially because they then head toward the exit with #44 leading the way and they were the first one get out through the exit after the device went off around time frame 374.
We believe that RFID #2, who mostly stays stationary near the exit, is a security person. Shortly before the explosion, at time frame 325, #44 arrives at the exit and decides to stay at the grid cell (70, 11), while #29 walks up to #2 at cell (72, 15) at frame 344, and stays there till the explosion. This leads us to hypothesize that #29 was trying to distract the security guard (#2) while #44 was in the process of remotely triggering the explosive device.
We believe the explosion happened in the office marked red in Figure 1 in the short answer page, since the two occupants of that room/cubicle were the first to die (they stopped moving before anyone else). Movements of RFID #21 looks suspicious as he was originally in the room right next the explosion area; he then entered the explosion room, and left shortly before the device went off.
After the explosion, #29 moves toward the corner (76, 2). #27, who stays stationary at the grid cell (71, 20) until the time frame 405 ends up following (chasing?) #29 and catches up with him. #17 who comes out of a room [grid cell (56, 17)] and arrives at the exit at time frame 476, seems to pause there for a moment and then also heads towards the corner (76, 2) and arrives there. It is possible that he sees a scuffle between #29 and #27 and decides to help. #38 (Loretta Middleton), who comes out of her room at time frame 404 also comes to the corner where #29 is with #17 and #27 already. We hypothesize that #29, as a suspect, gets caught and does not escape.
Most of the evacuees seem to escape near one of three locations: the corner near grid cell (3, 2); the corner near grid cell (89, 56) and the area near the grid cell (25, 2).
ID #59, who was the last person to die, was originally headed towards the exit near the corner (3, 2) but probably found the exit crowded, and reversed course toward the exit near the grid cell (70, 11). However, we believe this person came in contact with smoke or toxic materials from the explosion near the grid cell (40, 12) and died there at time frame 692. RFID #13, who was following #59, reversed course immediately to see #59 die and escaped to safety through exit near the corner (3, 2).
