NEMA traffic light stays red

[Nifemi954]

I have an issue with a simulation I am running, where the last vehicle does not complete the route but stops at a particular TLS. The simulation just keeps running afterwards.

Attached below is the config file;

<?xml version='1.0' encoding='UTF-8'?>
<configuration><input><net-file value="osm2.net.xml" /><route-files value="rsampler_trips900.xml" /><no-step-log value="True" /><time-to-teleport value="0" /></input></configuration>

[m-kro]

The config file does not contain any interesting setting regarding your problem. Does the vehicle enter the intersection or does it stop in front of the stop line? Sometimes a vehicle gets blocked when it does not have enough time to clear the intersection while other vehicles already enter it.
If this does not match your case, we might need the complete scenario files to watch the problem ourselves.

[Nifemi954]

Thanks for the reply. The last vehicle stops at the stop line before the TLS on red and the light doesn't change to green again.

Attached to this is the zip file containing a video showing the situation alongside the net, route and config file.

sumo.zip

[m-kro]

The problematic traffic light is cluster_181776620_8123131286_8893732102_8893737926_#3more (NEMA type). The last phase is not shown (traffic light doesn't switch states at all) although there is only one vehicle waiting for several minutes.

@qichaow @mschrader15 You know this code base best. Do you know the reason?

[mschrader15]

Had a look. The problem actually occurs more frequently than just the end of simulation.

Basically, because the NEMA controller only has phase 2, 6, 4 and 8, the detectors for the permissive left turn (phase 3, where the car waits) are not also being mapped to the corresponding main phase (phase 8).

The NEMALogic code only looks for detectors that correspond to lanes with "G" in the light string.

I'm not sure what the actual standard in this situation is, so as to encode in network autogeneration.

I will make a PR supporting defining multiple detectors per-phase, but might take some time

[Nifemi954]

Thanks for the reply, what would be the ideal way to write the Logic?

[mschrader15]

Below is how I would do it for the time-being. It serves 4 (controlling straight and left turn) and then 8.

Another option is to a put a minRecall on phases 4 and 8, so that they get served regardless of detection or not.

    <tlLogic id="cluster_181776620_8123131286_8893732102_8893737926_#3more" type="NEMA" programID="New" offset="0">
        <phase duration="90" state="srrGGgsrrsrr" minDur="10" maxDur="35" vehext="2.50" yellow="3" red="2" name="2"/>
        <phase duration="90" state="srrsrrsrrGGg" minDur="10" maxDur="35" vehext="2.50" yellow="3" red="2" name="6"/>
        <phase duration="90" state="GGGsrrsrrsrr" minDur="5" maxDur="15" vehext="2.50" yellow="3" red="2" name="4"/>
        <phase duration="90" state="srrsrrGGGsrr" minDur="5" maxDur="15" vehext="2.50" yellow="3" red="2" name="8"/>

        <param key="barrier2Phases" value="2,6"/>
        <param key="barrierPhases" value="8,8"/>
        <param key="detector-length" value="20"/>
        <param key="detector-length-leftTurnLane" value="10"/>
        <param key="minRecall" value="2,6, "/>
        <param key="ring1" value="0,2,4,8"/>
        <param key="ring2" value="0,6,4,8"/>
        <param key="show-detectors" value="true"/>
        <param key="total-cycle-length" value="80"/>

    </tlLogic>

[Nifemi954]

Thanks, I have made both changes and they worked.

[namdre]

Root problem remains to be fixed