Re: [sumo-user] Traci and sumo gui inconsistent simulation speed

["Tiago Furtado" <tiagofurtado@xxxxxxxxxxxxxx>]

Thank you for the response.  I believe I have resolved the issue. Just to give an idea of what I am trying to do. I am trying to have a distribution of traffic lights that communicate with each other over a network., with pairs of lights providing access to what should be a single bidirectional lane.  Upon the presence of a vehicle on either end, each pair of lights negotiates the priority of access based on a class of vehicle, heavy, light etc. 

I was using an external library to simulate the networking medium and having each node/tl have a separate thread as a message receive loop. Presumably issues with the Python GIL, or just servicing the various other loops was causing erroneous readings.

I was measuring time delays with the following code:

while traci.simulation.getMinExpectedNumber() > 0:

sim_step_start = getMillis()

traci.simulationStep()

sim_step_end = getMillis()

for i in trafficLightList:

i.lightState.run()

i.setLightUpdate()

step += 1

state_process_end = getMillis()

print(Fore.GREEN + "##########################################################")

print(f"[[{getMillis():.1f}] ] total sim:{(state_process_end-sim_step_start):.1f}")

print(f"[[{getMillis():.1f}] ] sumo sim: {(sim_step_end-sim_step_start):.1f}")

print(f"[[{getMillis():.1f}] ] state sim: {(state_process_end-sim_step_end):.1f}"+ Style.RESET_ALL)

presumably some other thread(s) were being processed between sim_step_start and sim_step_end as well as simulationStep(). Having switched to using internal python structures to simulate the medium the problem disappeared.  However as a note, I am using 1.19 and removing external control of TLS did make timing more consistent.  I still don't quite understand how the loop timings above could be shorter than the configured delay, regardless of whatever extra processing delay happens in the runner.py. 

On 9 November 2023 at 09:12:41 -00:00, Mirko Barthauer <m.barthauer@xxxxxxxxxxx> wrote:

Hi Thiago,

 

I cannot reproduce the problem with my simple example scenario. Is the timing more or less consistent if you do not call traci.trafficlight.setPhase? Do you use the most recent SUMO release? If you can confirm both, it would be great to have a minimal example scenario to reproduce.

 

Best regards

Mirko

 

 

 

-----Original-Nachricht-----

Betreff: [sumo-user] Traci and sumo gui inconsistent simulation speed

Datum: 2023-11-08T19:03:05+0100

Von: "Tiago Furtado via sumo-user" <sumo-user@xxxxxxxxxxx>

An: "sumo-user@xxxxxxxxxxx" <sumo-user@xxxxxxxxxxx>

 

 

 

 

Hi Everyone,

 

I am having an issue when using sumo-gui and a Traci with the simulation speed. Basically no matter what I set as the delay in the GUI I am getting a lot of cycle times that are both longer (acceptable for certain instantaneous steps where more processing is necessary), but I'm also getting cycle times that are much shorter than the delay. For example I'm measuring this in the runner.py , In the GUI I'm setting a delay of 200ms , and not only is the timing extremely inconsistent, it is not honoring the value I set. the value in sumo sim is measured in ms around  "traci.simulationStep()". 

 

##########################################################

[[24559.2] ] total sim:58.1

[[24559.2] ] sumo sim: 5.8

[[24559.2] ] state sim: 52.3

##########################################################

[[25248.3] ] total sim:231.8

[[25248.4] ] sumo sim: 102.8

[[25248.4] ] state sim: 129.0

##########################################################

[[25569.6] ] total sim:53.1

[[25569.6] ] sumo sim: 36.5

[[25569.6] ] state sim: 16.6

##########################################################

[[26019.1] ] total sim:2.6

[[26019.1] ] sumo sim: 1.0

[[26019.2] ] state sim: 1.6

##########################################################

[[26695.4] ] total sim:160.3

[[26695.5] ] sumo sim: 102.8

[[26695.5] ] state sim: 57.5

##########################################################

[[27190.4] ] total sim:134.8

[[27190.4] ] sumo sim: 77.4

[[27190.5] ] state sim: 57.4

##########################################################

[[27949.4] ] total sim:377.8

[[27949.4] ] sumo sim: 103.0

[[27949.4] ] state sim: 274.8

##########################################################

[[28124.0] ] total sim:22.7

[[28124.0] ] sumo sim: 21.5

[[28124.0] ] state sim: 1.2

##########################################################

[[28770.6] ] total sim:159.8

[[28770.6] ] sumo sim: 158.7

[[28770.6] ] state sim: 1.1

##########################################################

 

My script is subscribing to events from 13 different traffic lights and is controlling the phase of each of them via traci.trafficlight.setPhase(). Is there a way to make each step a more consistent time? Is it the number of lights I'm controlling? As far as I can tell the subscriptions aren't causing the variablility, but setPhase() seems to.

I would appreciate any insight, or suggestions of things to try.

Regards,

Tiago Furtado