certi-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: RE : RE: [certi-dev] Real-Time HLA simulation?


From: Eric Noulard
Subject: Re: RE : RE: [certi-dev] Real-Time HLA simulation?
Date: Tue, 29 Apr 2008 15:44:02 +0200

2008/4/29, Gotthard,Petr <address@hidden>:
>
>
> Where exactly do you see the conflict? I have a continuous simulation doing
> small steps (runs). After several runs, the HLA simulation time is advanced
> and the federate is frozen (waits) until simulation time = real time. If a
> time regulating federate don't advance the time, the whole federation is
> frozen.

I agree with this.
And moreover in a realtime simulation a federate (or a simulator) which
does not respect its realtime (linked to wall clock) constraints generates
"overrun".

As an example if you have a cyclic federate running at 10Hz if its
"small local steps" duration exceed 10ms it will overrun.

When you add HLA communication for linking several realtime simulation
you will have LocalStep_time + HLA comm_ time < max cycle time.

if your RT simulation is reactive (not cyclic) you have the same rule
but you have to measure HLA round-trip communication in order
to see if it fits your RT latency requirement.

___________________________
>  From: On Behalf Of HADJ AMOR HASSEN
>

>
> My problem is how to mix Real Time algorithms (waiting for real time) with
> HLA time management services. A conflict will happen for sure.

Nope.

1) do your local computation
2) SendUpdate with current time
3) TAG to whatever time you want

> Let be a federate1 which updates its attributes at t=t1. with the time
> elapsed during the data transfer, the other federates will receive the new
> values at t=t1+t'. So the simulation isn't real time.

This is a general distributed simulation issue which is not HLA-related.
If an event occurs in some place you need to take some time
to "observe" the event some where, just like storm and sound.

the "only" problem with HLA is that you do not have a precise
"communication duration time".

> I am also thinking of another possibility: Can we try to freeze all the
> simulation (federation) to make it real time without carrying if federates
> are real time or not.

freezing may simply be calling TAG and waiting for the RTI to say YES.

-- 
Erk




reply via email to

[Prev in Thread] Current Thread [Next in Thread]