Re: [sumo-devel] request for comments on maximum speed extensions

[Daniel K. <d.k...@go...>]

Hello all,

ok, I now added the following for speed factor computation:
    const SUMOReal devA = MIN2(SUMOReal(2.), MAX2(SUMOReal(-2.),
rng.randNorm(0, 1.)));
    return (devA*mySpeedDev + 1.) * mySpeedFactor;

The rng was moved to MSVehicleControl and now serves within choosing
routes and vtypes from according distributions as well as for
computing the speed factor. I suppose that's all for this topic, isn't
it?

sincerely,
Daniel

2012/6/13 Michael Behrisch <beh...@us...>:
> Hi Daniel,
>
> 2012/6/8 Daniel Krajzewicz <d.k...@go...>:
>> b)
>> Ok, I implemented speed deviations by calling an MSVehicleType method named
>> "computeChosenSpeedDeviation" in the MSBaseVehicle and setting this value
>> as lane max. speed factor for the vehicle. It is currently computed as:
>>
>>    SUMOReal devA = MIN2(+2.*mySpeedDev, MAX2(-2.*mySpeedDev,
>> mySpeedDevRND.randNorm(0, mySpeedDev)));
>>    return devA + mySpeedFactor;
>>
>> Could someone verify this?
>
> The specification
> http://sumo.sourceforge.net/doc/current/docs/userdoc/Specification.html#Types
> says we apply the speed factor first and then the deviation factor. So
> if we have a lane speed of 50, a speed factor of 2 and a deviation of
> 0.1, we should have a normal distribution with mean 100 and std dev of
> 10. With the code above we get mean 100 and std dev 5. If it fits user
> expectations better we can change the spec as well.
> Furthermore I am not 100% sure about the randNorm implementation. It
> calls the parameter "variance" but the way it is used looks more like
> standard deviation. To be on the safe side we could simply:
> const SUMOReal devA = MIN2(2., MAX2(-2., mySpeedDevRND.randNorm(0, 1.)));
> return devA*mySpeedDev + mySpeedFactor;
> or to be spec compliant:
> return (devA*mySpeedDev + 1) * mySpeedFactor;
>
>> I introduced an own MSVehicleType RNG, initialised in sumo_main.cpp - you
>> can now pass a pointer to RandHelper::initRandGlobal (should be probably
>> renamed).
>> This means all RNGs will use the same seed/methodology as the usual static
>> one. I think that's fair enough.
>
> My intention was to solve
> https://sourceforge.net/apps/trac/sumo/ticket/444 so we should use
> this rng also for choosing from the vtype and route dists (and rename
> it accordingly :-)).
>
>> Now, a lane has no longer a getMaxSpeed-method, but a method named
>> getSpeedLimit which returns the maximum allowed speed, and a method named
>> getVehicleMaxSpeed which obtains a vehicle and returns the speed the
>> vehicle wants to use on this lane. I assume that this is not the fastest
>> solution, but I think the most clean.
>
> I like it,
> Micha