Hi,
Just recently we did some traffic/evacuation simulations for indoors. In
particular we have utilized the multi-agent traveling simulator MATSim
(www.matsim.org) and IndoorOSM
(http://wiki.openstreetmap.org/wiki/IndoorOSM) data for our simulations.
The scenario is based on a university building which is fully occupied,
resulting in 313 agents. Some results are combined in a small video:
http://www.youtube.com/watch?v=7u9g022IrVY
Greetings,
Marcus
Am 03.04.2012 12:38, schrieb sourav dutta:
Hi,
Sorry for mailing in so late, i came to know about Gsoc few days ago.
I am Undergraduate doing my Btech-Hons for IIIT - Hyderabad. I have
worked in OCR, Vision,(Sfm) ,Image Processing,Information retrieval.
OSM data have been used for path finding...with A star algo etc.but it
is never used for simulation of traffic.
Traffic monitoring and simulation has been worked on for some time now.
There are very elaborate systems like MITSIM which gather data from
various sources and design a traffic model. But the problem is it is
very difficult to implement them. The information about the streets,
traffic
data, satellite data etc. is difficult to gather.
My idea is to design a traffic simulation using OSM data. as the
OSM data is readily available and is available in many formats. To model the
traffic probabilistic or network flow models are popular. But to get a
more accurate simulation want to use multiple vehicles as bots
interacting with
each other.
A basic overview of my idea ...
1) first we need the area where want the simulation to run. This would
be done by defining a rectangular region in the map. This data can be stored
in posgreSQL which will make it easier to use the data.
2) Next we define for all the bots - source and destination and this
find the appropriate path(A star) to go.
3) At the heart of the simulation we need to have a engine which would
handle the collision etc. and directs the bots etc. I have tried a few
physics
engine but their performance degrades steeply decreases with the
increase of number of objects.
4) So I want to use my own engine with only few rules for collision
etc. The brute for implementation would require to check each bot with
other bot
( the same thing which makes other engines slow O(n^2) ). But i plan
to use a implementation of KD-tree, ANN(Approximate nearest neighbour) Open
source implementation of the same exists and works pretty well. This
would allow the matching to be done in O(nlogn).
ANN implementation (http://www.cs.umd.edu/~mount/ANN/
<http://www.cs.umd.edu/%7Emount/ANN/>)
5) To manage the interaction of the bots the memory requirements
skyrockets as the number of bots increases. ie why i wish to treat them
as Multi-Agent-System
There are nice implementations of "distributed multi-agent
framework" which use shared memory to manage the agents( bots in our case).
6) congestions , bottlenecks and collisions can thus be detected. In
the streets.
7) Future Works - Once this frameworks is complete data from other
sources like the traffic lights flyovers etc.. can be incorporated in
the existing system.
--
Sourav Dutta
CSE,UG3
IIIT H
_______________________________________________
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk
_______________________________________________
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk
