Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [sumo-user] Perception errors
1. The error state evolves according to  https://en.wikipedia.org/wiki/Ornstein%E2%80%93Uhlenbeck_process
The code is in MSDriverState (see OUProcess)
Control parameters for the OUProcess are errorTimeScaleCoefficient and errorNoiseIntensityCoefficient as well as awareness
(see https://github.com/eclipse/sumo/blob/7ca399daab28139f5a726f557c4593454cb5174a/src/microsim/MSDriverState.cpp#L180-L188)

2) there is no range limitation

3)
a) the error affects gap and speed difference estimation (by default the error coefficients have the same sign which means a driver that overestimates the gap to the leader will also overestimate the leader speed)
b) the underlying assumption was that the expected error is 0 (no systematic bias). For the calculation see answer to 1)
c) with errorState = 10 the actual difference between real gap and faulty gap is 7.5 * trueGap (with default headwayErrorCoefficient = 0.75)

Note: when doing a sample run of the OU-Process a million times with default parameters, the error lies in [-0.85, 0.85] and ~90% of the error states are in [-0.2, 0.2].

regards,
Jakob

Am Di., 26. Jan. 2021 um 08:20 Uhr schrieb Mehdi Maleki <mmehdi@xxxxxxxxxxxxxxxxxxx>:

 

Hi,

 

Regarding the driver state device:

 

 

 

 

  1. How is the error(t) scaled, do we give input for this? Is this the same with device.driverstate.errorState?

 

  1. Default value for speedDifferenceErrorCoefficien is 0.15, what is the range for this?

 

 

 

Before, I asked about the definition and range of the "device.driverstate.errorState" parameter. The answer that I got was this: ErrorState gives the magnitude and direction of the relative error in gap estimation. The state tends toward 0 in the long run but there is no limit to how far it may deviate from 0.

 

  1. My questions are: Does it affect only the gap estimation? Why it tends to zero over the time, does it represent something in real life? With what rate it tends to zero? (I want to see the mathematical calculation, for example when I set errorState to 10 how much the real gap and faulty one are different).

 

Btw, I have checked "MSDriverState" file but could not find the answer for my questions.

 

 

 

Best regards,

Mehdi

_______________________________________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user

Back to the top