Two types of research are going on:
i. Developing a simulator from the scratch
ii. Tuning up an existing simulator MITSim
We have tried our best to tune up MITSim. But at last we have come to a conclusion that we should build up our own simulator. Because Our traffic phenomena are different than others. Our traffic system is heterogeneous and non-lane based. We have non-automated vehicles and the main constraint is our driver don't obey traffic rules.
We have already build up our own simulator "DhakaSim" in JAVA and C++.
DhakaSim:
DhakaSim mimics real traffic scenarios in a virtual environment using real road network topology and real traffic data. The purpose of our project is to evaluate alternative transport options in Dhaka city.The constructed simulator receives network topology, driver behavior, and origin-destination flow as input. The simulator generates vehicles for each origin-destination pair based on the provided parameters, visualizes the flow of vehicles, and provides statistics. The visualization allows decision makers to identify different traffic events visually and the statistics allows them to analyze the events.
Current work:
Now we're trying to make our simulator distributed because of huge amount of computation. We have chosen OPEN-MPI to make our system distributed.
So We need to partition our road network in K region where K is the number of machine and in each partition number of node must be balanced.
This problem is known as popular problem K-minimum cut which is NP complete problem.
Now we're trying to develop an approximation algorithm to solve this issue.
Suggestions:
We have received several suggestions. These are as follows:
1. How historical data will be processed? Probability distribution??
2. Incorporate Non-Ideal Phenomena (such as accident, horn, signal).
3. Automatic Traffic signal suggestions.
4. Automatic signaling to vehicles ( VANETs)
5. Incorporate Driving behavioral model (Speed, Acceleration, signal, horn, inertia etc) .
6. Synchronization problem on distributed computing.
7. How will be the partition of road network? Link-wise or Node-wise??
8. Intersection modeling.
9. Vehicle list shuffling or subset of vehicle list shuffling.
10. Routing problem. ( Force vehicle to follow certain link )
11. Traffic police modelling.
12. Real Data collection.
13. Follow poison distribution.
14. Model the main branch first. (Shahbag to Farmgate)
15. It should be web-based.
16. It should be build not only for dhaka city but also for our whole country.