| [sumo-user] Questions about TraCI performance |
|
Dear sumo-Users, i am currently interfacing SUMO (v.032, official release) via the TraCI client within C++ by subscribing to the vehicle states as well as triggering SUMO to step forward with the time. At the same time, SUMO (we are using the server without the GUI) tells us, how much time SUMO actually needs: >> Step #235.64 (0ms ?*RT. ?UPS, TraCI: 2ms, vehicles TOT 82 ACT 51
>> Step #235.77 (0ms ?*RT. ?UPS, TraCI:
27ms, vehicles TOT 82 ACT 51 >> Step #235.90 (0ms ?*RT. ?UPS, TraCI: 2ms, vehicles TOT 82 ACT 51 >> Step #236.03 (1ms ~= 10.00*RT, ~51000.00UPS, TraCI: 1ms, … >> Step #236.16 (0ms ?*RT. ?UPS, TraCI:
27ms, vehicles TOT 82 ACT 51 The point is, that sometimes it seems that the real time factor cannot be determined appropriately and also the TraCI time seems to jump between two steps. What is really confusing, is the fact, that we are measuring
the time, which is necessary to tell TraCI to step ahead with SUMO. This time is about 5 to 10ms, depending on the number of vehicles administrated by SUMO and subscribed by our application. So what I am interesting in, are the following two questions: ·
Why does the TraCI time, given by sumo server, is sometimes jumping, although we are measuring an almost constant time in our external application for stepping forward with TraCI?
·
I mean, there are a lot of different parameters (scenario, pc configuration, subscribed vehicle quantities, number of vehicles, update rate) to given a precise indication, but are there any experiences of how
many vehicles can be simulated and transmitted by TraCI to reach real-time? Best regards, Marc |