[Uni-verse] Uni-verse test session 25-27/4 2004

[Peter Lunden]

Uni-verse test session 25-27/4 2004

Present: Peter Lundén, Raine Kajastila, Mattias Claesson and Akos Schreck

The goal of the test session was to establish a stable demo setup. Most 
Uni-verse software has problems with more complex models so we have 
chosen to use a quite simple model, a sightly modified version of 
testscene model, the modifications consists of removing the doors and 
the door frames. The model is found at: 
http://www.pdc.kth.se/~slas/universe/models/testscene_no_doors_3.vml 
<http://www.pdc.kth.se/%7Eslas/universe/models/testscene_no_doors_3.vml>

The model was scaled by a factor 0.4 before uploading it to the server.

Two sound sources where added to the scene, one in north west (north is 
at negative z) corner of the hall and the second in the north end of the 
corridor.

The sound renderer (UVSR) uses a headphone configuration.

We did only test the OSGRenderer because there where serious problems 
with Quelsolaar during the last test session.

Setup 1:

3DSMax----------------
                                   |
OSGRenderer---- verse ------- UVAS -------- UVSR
                                   |
Material DB ----------

Setup 2:

3DSMax----------------
                                   |
OSGRenderer ---- verse ---- reducer ---- verse---- UVAS ---- UVSR

Testing with setup 1:

* Navigating from OSGRenderer work, although is still steals the object 
with id 0.

* Scaling geometry in 3DSMax was correctly reflected in the whole chain, 
although OSGRendnerer was reacting very slow and BSP structures in UVAS 
was degenerated after many scalings.

* Creating and moving a sound source from 3DSMax was working.   

* Creating and changing materials from 3DSMax was working.

* Errorous transmissions did infrequently occur in UVAS. 

* Long delay before the refections of the second sound source gets 
calculated.
 

Testing with setup 2:

* Reducer affected geometry in strange ways, IE the floor was raised 
about 75 cm. 

* The reducer does not reduce in the sense UVAS needs. Number of BSP 
cells in UVAS was 111 in the unreduced model and 225 in the reduced.

* Problems with material handling, the problems was partly fixed during 
the session. 

* Material handling problems stop us from testing more complex models.

OSGRendnerer was updated during the session. A new problem occurred 
after updating; part of the geometry was now missing. 

Conclusions:

We can get a demo that works at the review but more work need to be done 
on most software components to increase stability and in some cases to 
fix its functionality.

The testscene model can work as a demo model. We propose that the model 
is made more visually impressive by adding objects like furnitures 
(without acoustic materials) and by adjusting the visual materials. We 
need more than one model that can be used in the demo. A slightly more 
complex model should be developed.  

The geometry reducer did not work with UVAS, it should be redesigned to 
work better with UVAS.  The reducer should not make any assumptions 
about the order of how applications are connected and when the model is 
uploaded to the server.

More work is needed on the visual renderers, we need both OSGRendnerer 
and Quelsolaar to work at the demo.

--PLu