[Opensim-users] objects undergoing continuous change
Paul Fishwick
fishwick at cise.ufl.edu
Wed Nov 12 16:16:51 UTC 2008
Crista:
Diva Canto wrote:
> I have written a region module that does traffic simulation. Here's a
> video produced by my students:
> http://uk.youtube.com/watch?v=kJNDcurLP1w
This very impressive and nice use of OpenSim along the lines of what I
was thinking about. For teaching discrete event simulation, the entire
messaging
passing approach seems to lend itself well to this (assuming FIFO
ordering on event queues for a region). For continuous simulation,
your vehicles appear to be updating position at a decent frequency.
Did you use llSetPos() and was there a delay or did you run flat-out
at whatever the server could muster?
For some reason, my simple test was not good, perhaps because I was
moving too short a distance during updates. I'll retry again later tonight.
>
> That was back in late August; the opensim API has changed considerably
> since then and I haven't had time to update my module yet. I didn't
> use physics; it was all done with periodic updates to the objects'
> state, just like in traditional simulation. It works pretty well, only
> starting to break when there are more than 2 avatars in scene; the
> breakage is because we have the server behind a home DSL, so packets
> start getting lost.
Yes, I could see how too many updates could be a problem depending on
the server
and location of clients.
>
> In general, for doing tradional simulations forget about using inworld
> scripting. Even if the scripting engine was already working well,
> which is not, that introduces a lot of unnecessary overhead.
I suppose I was wondering whether the overhead could be removed or
minimized for
running standalone servers (or at least, servers with no more than 1 or
2 avatars)...
> Moreover, it also imposes a programming model that is an obstacle to
> simulations (for example, not being able to hold references to objects
> and having to interact with them via llSay functions).
Yes, having the option for direct referencing would be useful.
> The opensim API, on the other hand, is a joy to work with (how many
> times have I said this? :-)
When you say the opensim API are you referring to LSL, the C# interface,
or something else?
I've been using LSL, but haven't seen any docs on the C# API yet.
-p
>
> Diva / Crista
>
> Paul Fishwick wrote:
>> Justin Clark-Casey wrote:
>>
>>> Yes, things are stored in memory with a regular flush to database that occurs every 15 seconds. However, I suspect that
>>> objects with the physics attribute are not currently persisted (from a recent thread on opensim-dev).
>>>
>>>
>>>
>>
>> I am curious as to whether it could become possible (in a standalone
>> server or
>> in a grid with only one person) to do continuous visual updates for moving
>> objects (physical or otherwise). I have tried a few experiments, such as a
>> continuously moving block using LSL (setpos) and after a few seconds, the
>> behavior becomes erratic (for example, the box shoots off somewhere or
>> disappears entirely from the region).
>>
>> I realize that having fast moving visual objects does not lend itself
>> well to sharing
>> state, but there are some instances where one wishes to simply launch a
>> region
>> or grid and do more traditional simulation (as one might do in OpenGL)
>> whether
>> or not the object states are shared, or on the frequency of sharing. It
>> would be nice
>> to have this "standalone continuous object motion" as a switch-based or
>> deltatime-specified capability.
>>
>> Thoughts?
>>
>> -p
>>
>>
>>>> * Would it (or is it) useful to have some scripts run on the client
>>>> rather than the
>>>> server for less server loading? For example, could ODE run on the
>>>> client
>>>>
>>>>
>>> I think that some minor stuff is done client side already (such as flexiprims). However, I suspect one runs into
>>> difficulties when objects can affect each other, since you would have to start communicating all this information back
>>> to the server and somehow synchronizing it with changes from other clients.
>>>
>>> However, I am not an expert in this area.
>>>
>>>
>>>
>>
>>
>>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Opensim-users mailing list
> Opensim-users at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-users
>
--
Dr. Paul A. Fishwick E-Mail: fishwick at cise.ufl.edu
Dept. of Computer & Info Phone & FAX: (352) 392-1414
Science and Engineering WWW: http://www.cise.ufl.edu/~fishwick
University of Florida (PGP Key available at above WWW address)
P. O. Box 116120
332 Bldg. CSE, Gainesville, FL 32611-6120
More information about the Opensim-users
mailing list