Hi,
We have developed a parallel framework over SUMO. When I am copying vehicles from one partition to another. With the vehicles I am also copying the following lane change parameters. I am keeping sigma 0, --random false and same seed value for all the partitions and sequential SUMO.
"laneChangeModel.speedGainProbabilityRight"
"laneChangeModel.keepRightProbability"
"laneChangeModel.lookAheadSpeed"
After adding vehicles in the next partition, I am also moving them for the desired position in the current step and setting speed for the current step.
i am calculating the current step speed as follows:
double vthis=LastTimestepSpeed+ACCEL2SPEED(LastTimestepAcceleration);
Or if the vehicle has a Leader...then with the follow speed.
But if I compare travel time of this parallel framework with sequential SUMO, I am getting an accuracy error.
I have found that patchSpeedBeforeLC is calculating some different speed which is causing randomness.
vDawdle = MAX2(vMin, dawdle2(vMax, sigma, veh->getRNG()));
This randomness calculation should be off when we use --random false.
Thanks,
Mahima