Multiple output analysis

All topics related to questions in Pedestrian Dynamics
RiRao90
Posts: 5
Joined: Saturday 19 November, 2016 - 15:29

Multiple output analysis

Postby RiRao90 » Monday 27 February, 2017 - 18:44

Hello,

I've got a question regarding the output analysis.
Quick example:
Let's say I just simulated 10 replications. Now, I want to select from each replication the provided escape time to aggregate this time. If I have additionally 20 different scenarios to choose from that would be 200 replications to select from by hand. I don't really want to pick up this much data manually.

So my question is:
It seems PD only allows me to export data from only one replication at a time (as csv or something else). I did not find any function for multiple replications. Am I missing something or does PD not support multiple data selection respectively data export?


And a secondary issue but this is not so important to me. I want to change the parameter of the number of server queues (e.g. from 50 to 75 server queues) within an activity after a couple of scenarios while doing an experiment. PD has a problem with it, the agents are stuck after the change and are not leaving the activity. I have to destroy the atom so the agents can leave the activity again. i did not find a solution why this happened. I tried to update every single statement but I guess the problem is how I am setting the QueueStrategy?
I cannot provide a screenshot or error message since I have PD running currently but I can change the server queues manually.

Best regards,
Ralf

User avatar
marlies
Posts: 288
Joined: Monday 17 January, 2011 - 09:28

Re: Multiple output analysis

Postby marlies » Tuesday 28 February, 2017 - 10:39

Hello Ralf,

There is a function that allows you to analyze the general statistics of multiple replications. You can use it calculate a.o. avg/min/max for the number of agents and travel time statistics. This function should make it possible to calculate the evacuation time over a number of replications.

To use this option you'll have to do the following:
- Open the interact window (Tools > 4DScript interact or Shift + F6)
- Use the function:

Code: Select all

AgentStatistics_CalculateOverallOutputStatistics( [C:\My Documents\Pedestrian Dynamics\3.0 Official Release\PD_Results\Experiment1\], STATISTIC_average, 0, 1, 5 )
Where:
p1: the path of the folder containing all runs for the experiment. Note that the path must be between square brackets
p2: the statistic type that you want to have. Available options are: STATISTIC_CONFIDENCEINTERVAL (one-sided), STATISTIC_PERCENTILE, STATISTIC_AVERAGE, STATISTIC_MAX, STATISTIC_MIN.
p3: statistics value (when analyzing confidence interval or percentile)
p4: replication number first run
p5: replication number last run

Executing this script in the interact window will give you a window with an overview that can be copied and pasted to Excel.

Your second question about updating the number of server queues dynamically can be done in different ways:
- Update the number of server queues on reset. However this will require some scripting as these server queues are internal elements that need to be created and connected in the right way. If you want I can provide the code to that, however you could also:
- Set a maximum to the number of server queues that you won't exceed. Then use only the number of server queues that are available according to your scenario setting. You can do that by changing the setting Queue Strategy to something like this (assuming you have table scenario settings with the max number of server queues to be used):

Code: Select all

{**Smallest queue**} IndexMin(GetScenarioSettings(1,1), Content(in(Count, c)))
See also attached example.

Regards, Marlies
Attachments
20170228 Server Queues.mod
(162.43 KiB) Downloaded 19 times

RiRao90
Posts: 5
Joined: Saturday 19 November, 2016 - 15:29

Re: Multiple output analysis

Postby RiRao90 » Wednesday 01 March, 2017 - 09:44

Hi Marlies,

I'll try the provided solution for the analysis. This seems straightforward so I hope I don't get stuck there.

Regarding my other issue. I have a table where my settings are stored but I guess the problem was to set the number of server queues way higher since I always reached my limit. I will give this another shot with your second provided solution.

Thanks and best regards,
Ralf

RiRao90
Posts: 5
Joined: Saturday 19 November, 2016 - 15:29

Re: Multiple output analysis

Postby RiRao90 » Wednesday 08 March, 2017 - 16:10

Hi Marlies,

about the function to analyze the general statistics of multiple replications I came across a problem (see attachments). It seems only some replications have this problem because in a range of let's say 1-5 in Scenario 1 the function works fine. If I add 6 to this range this error appears. 7-9 might work, 10 again doesn't work and so on. In Scenario 2 other replications are working, others might not.
Also replication 1-5 equals replication 10 to 14 (from the arguments e4, e5) when I am running the function which is kinda strange. 10 is then replication 18 and so on. I guess this not a big problem?

So is there some workaround for the TImeStr or are replications buggy? Maybe I could exclude the buggy ones but since I can only start and end with replications this would be a bad solution. Not to talk about that I don't really want to exclude any data.

Best regards,
Ralf
Attachments
PD Error.PNG
PD Error.PNG (5.49 KiB) Viewed 292 times

User avatar
marlies
Posts: 288
Joined: Monday 17 January, 2011 - 09:28

Re: Multiple output analysis

Postby marlies » Wednesday 08 March, 2017 - 16:38

Hi Ralf,

It's a bit hard to say what exactly causes the problem without the data. So my question is:
- Is it possible to send me the model so I can try to reproduce (I'll send you a PM for contact details)?
- I would also like to know which statistic type & other parameters you used when this problem occurs?

I'll also have a look at the replication numbers. It looks to me as if the names are not updated correctly when you try to analyze different subsets in a row.

Regards, Marlies


Return to “Modeling in Pedestrian Dynamics”

Who is online

Users browsing this forum: No registered users and 1 guest

cron