A quick blog update to play with a neat agent simulation tool integrated into OpenStreetMap mapping tools with vehicle driver behaviour modeling.
SMARTS – Scalable Microscopic Adaptive Road Traffic Simulator (University of Melbourne)
At a Glance
I ran the JRE8 simulation on an OSM map of Mahidol University, Salaya district (downloaded from openstreetmap.org) with 10,000 agents and 1 hour duration, without bots taking alternative routes in the event of congestion.
However, the agents prefer the park and the palace instead of the city commute. Below is a screenshot example after 11 minutes. After 1 hour, the agents continued to converge on the already congested roads.
On the same OSM map of Salaya (downloaded from openstreetmap.org) with 20,000 agents and 12 minute simulated duration, this time bots reroute if congestion is encountered.
In this trial, the number of congested roads increased and the queue lengths increased. The rerouting and the increase in agents likely have contributed to (firstly) the increased number of congested roads and (secondly) towards queue lengths.
As a matter of subjective interest, the congested roads seem to more closely match my existing expectation / experience driving on these roads. This part of Bangkok has some impressive traffic jams!
Future Actions to make the simulation more effective:
the commuters need scripting from origins (housing estates,etc) to destinations (often the inner city).
- separating individual intersections for more realistically deconstructed modelling.
This post brings together a couple of trials using the agent simulation software released by the University of Melbourne SMARTS team.
Though it’s a far cry from a comprehensive review. It was fun to play with and I can imagine, that with some work, a visualisation can be generated to support Transport Policy Recommendations for Road Network changes (or redesigns). If you’re interested you can visit the SMART Project website or get in touch if you fancy discussing possibilities.