faq 111991329
by Abdullah Binthunaiya on 2017-09-13 15:11:39
I am simulating 150,000 agents from origin points to destination points using Dijkstra algorithm in 2 scenarios: (1) single mode and (2) multimode. Results for the car mode are unrealistic and they are affecting the other scenario. In scenario (1), I have the following issues: a- Car Users stay in the network for very long time; What could be the reason for that
b- I saw executed plans that are of
leg type "pt" although only mode in the Network XML is "car".
c- No scoring graph was generated, unlike the case when I un multimode scenario.
d- I could not find a way to see the statistical distribution of selected plans for all agents, I mainly need to get : Travel Distance and Travel Time. I can see these for individual agents in output plan but it i hard to go over the all 150,000 plan and report them individually.
e- I also want to visualize the executed plans (route or Links) for every agent on open street map or ArcGIS, is there a way to do that. (write them to .osm or .shp)
by Kai Nagel on 2017-09-14 10:38:26
You need to give your "true" activities (such as "home" and "work") a typical duration (e.g. 12:00:00 and 8:00:00 ... they do not have add up to 24:00:00).
Let us know how it goes after that.
by Abdullah Binthunaiya on 2017-09-15 16:21:19
Thanks a lot Dr. Nagel, There is an improvement "pt" disappeared, and scoring is now calculated and graphed.
However, I see many agents not arriving although they departed , speed on all links is 22.222 (80 Km/hr) and many links have more than 2 lanes. 150,000 agents are supposed to leave 30 work locations to 30 residential areas locations. In my study area, this number is not too large in a typical working day, and trips should not exceed 1 hour at most (for the locations I selected). attached the log file and the config file. may you please have a look at it and share what you think is not set correctly....
by Kai Nagel on 2017-09-15 20:05:05
Re "analysis". For the time being we recommend to write your own events handlers. There is some material in the tutorials, or parse the events file by some other means (e.g. in python if you like that better). There is some analysis code that does what you need, but it is in a contrib, and it will just make your life harder.
by Abdullah Binthunaiya on 2017-09-20 17:20:05
Thanks a lot Dr. Nagel, The problem was solved when I reduced the number of agents per location. I also fixed some of the parameters but I believe major issue was the number of agents, they depart from nearby locations and they are 5000 per location that they have to queue up for long time