faq 166395927

JDEQSim vs Qsim

by Pradeep Burla on 2018-02-04 21:21:58

---

Hello everyone,

I am currently trying to understand the actual difference between in terms of the simulation output. Theoretically, Qsim is time step driven whereas JDEQsim is deterministic event based. I expected the simulation results to be consistent with each other. If I increase the delta T(10 seconds) in qsim I expect the results to be coarser than for a lower delta T (1 second). And lower the deltaT, the results should tend closer to JDESQsim results. But this does not seem to be the case.
With delta T =1 second, all the agents complete their activities in Qsim. The same simulation with JDEQSim shows a lot of agents are still en route. This does not change until I reduce the "Squeeze time" parameter from 1800 seconds to 10 seconds. Inspite of the change, still the results are a little different.
Is it due to congestion, if yes how does Qsim handle congestion.

Would be great if someone could explain me this behavior.

Thanks in advance :)

Regards,
Pradeep

---

Comments: 2

---

Re: JDEQSim vs Qsim

by Kai Nagel on 2018-02-04 23:30:20

Setting the "squeeze time" (called stucktime in qsim) to the same value in both mobsims is certainly a good idea.

The simulations are not meant to produce the same results. If anything, one could say that with timestep➝0 (!!), qsim should approach jdeqsim. Yet in reality, I still would assume that they generate different results, because of smallish(?) implementation details.

Still, it is hoped that they generate similar congestion patterns when run with similar parameters (e.g. the squeeze/stuck time; presumably set qsim to "kinematicWaves and make sure that the wave backpropagation speed is the same between the two, etc.). I have always found that similarity when I ran both. Not sure if anybody ever published a comparative study. Might be a good contribution to the matsim user meeting.

You probably would have to go through papers and code for details.

---

Re: JDEQSim vs Qsim

by Pradeep Burla on 2018-02-05 00:56:44

Dear Prof. Nagel,

Thank you for your prompt reply.
I did try set the squeeze time and stuck time to similar values and the results seem promising. The results match much better. I believe I need to calibrate the other parameters to get consistent results.

To add on just as a note, there is a difference in the way an agent travels in QSim and JDEQSim. I have attached an image as an example.

Agent in qsim travels in total 3500 m, whereas agent in JDEQsim travels 4000m. The links assigned to the activities are same in both the mobility simulators, but link travel behavior changes.

Also thanks for the suggestion on a comparative study between both the mobility simulators. Would post something if I make progress on it.

Regards,
Pradeep