The only way to get this kind of delay is via an external (traci) control
script that (continuously) records the speeds desired by the model at time
t and overwrites the speed at time t + 2. See here for a related discussion
and supporting traci functions: https://github.com/eclipse/sumo/issues/7190

Am Sa., 5. Dez. 2020 um 00:11 Uhr schrieb Yasmina Jaafra <
[email protected]>:

> Hello Jakob,
>
> Ok, now I understand better the action-step-length option. However, I need
> to reflect a delay in action application while training the reinforcement
> learning agent. Indeed, in real-world a brake decision at t (k) is executed
> at t (k + 2) due to the vehicle transmission mechanisms which require some
> time to transmit the driver maneuver until wheels. Is there any
> functionality in SUMO that can support such training?
>
> Best regards,
>
> Le ven. 4 déc. 2020 à 12:30, Jakob Erdmann <[email protected]> a
> écrit :
>
>> No. It means that vehicles will only change their acceleration at most
>> every 0.3s and will keep constant acceleration for 2 out of 3 simulation
>> steps. Thus their reaction to certain events will appear delayed. Since
>> each action is executed as soon as it is computed (in an action step) there
>> is no queue.
>>
>> Am Fr., 4. Dez. 2020 um 13:07 Uhr schrieb Yasmina Jaafra <
>> [email protected]>:
>>
>>> Hello Jakob,
>>>
>>>
>>> Thank you for your reply.
>>>
>>>
>>> I enabled the "–*default.action-step-length option". Currently I have
>>> these settings:*
>>>
>>>
>>> *step length = 0,1 s*
>>>
>>> *action length = 0,3 s*
>>>
>>>
>>> *Does this mean that an action taken at time step t(k), will only be
>>> executed at t(k+3)?*
>>>
>>> *And given that Python client is sending an action every time step,
>>> would there be an action queue on Sumo side to manage the action control
>>> delay?*
>>>
>>>
>>> *Regards,*
>>>
>>> Le jeu. 3 déc. 2020 à 15:29, Jakob Erdmann <[email protected]> a
>>> écrit :
>>>
>>>> see https://sumo.dlr.de/docs/Car-Following-Models.html#actionsteplength
>>>>
>>>> Am Do., 3. Dez. 2020 um 14:29 Uhr schrieb Yasmina Jaafra <
>>>> [email protected]>:
>>>>
>>>>> Hello,
>>>>>
>>>>> I would like to apply control action delay on the ego vehicles. Is
>>>>> there any option to enable the delay between states and actions in Sumo.
>>>>>
>>>>> Best regards,
>>>>>
>>>>> _______________________________________________
>>>>> sumo-user mailing list
>>>>> [email protected]
>>>>> To unsubscribe from this list, visit
>>>>> https://www.eclipse.org/mailman/listinfo/sumo-user
>>>>>
>>>> _______________________________________________
>>>> sumo-user mailing list
>>>> [email protected]
>>>> To unsubscribe from this list, visit
>>>> https://www.eclipse.org/mailman/listinfo/sumo-user
>>>>
>>> _______________________________________________
>>> sumo-user mailing list
>>> [email protected]
>>> To unsubscribe from this list, visit
>>> https://www.eclipse.org/mailman/listinfo/sumo-user
>>>
>> _______________________________________________
>> sumo-user mailing list
>> [email protected]
>> To unsubscribe from this list, visit
>> https://www.eclipse.org/mailman/listinfo/sumo-user
>>
> _______________________________________________
> sumo-user mailing list
> [email protected]
> To unsubscribe from this list, visit
> https://www.eclipse.org/mailman/listinfo/sumo-user
>
_______________________________________________
sumo-user mailing list
[email protected]
To unsubscribe from this list, visit 
https://www.eclipse.org/mailman/listinfo/sumo-user

Reply via email to