Dear Michael, and others,

I will take your advice and experiment with this modification of the
KeyNavigator,
however this is not our principal device. I actually detected the problem
with the
perspective projection in Java3D while I was wearing a see-through HMD
tracked
by an orientation tracker and a position tracker. These devices have been
calibrated
are reporting resonable results. Well, when I turn my head, so that I always
see
the real counterpart of 1 and 2 aligned, It is not the case with Java.
However, like I
said before, a rotation only is applied to the viewplatform. So this problem
is not
only related to the KeyNavigator.

I can attach a picture if needed to understand the problem. Please try to
experiment
with that and you will see. Place a cube in the world (let say in 0, 0, 0
and of size
2 on any side). and place the viewplatform in (1, -5, 0) so that one side of
the cube
appears as a line. If you only rotate the transform applied to the
viewplatform then
you should always see this side as a line. In my case it does not work this
way,
I see the side as I turn.

Any idea?

Thanks

Yohan

>Date:    Mon, 4 Oct 1999 16:51:38 -0700
>From:    "Michael F. Deering" <[EMAIL PROTECTED]>
>Subject: Re: PERSPECTIVE COMPUTATION, WRONG???
>
>There is a "bug" in the current KeyNavigator.java code, depending on how
>one wants to define the semantics of rotation. While a more complete
>re-write is underway, a single line change we found may give the
>behavior you want.
>
>Change the line:
>   vpQuat.mul(lrQuat, vpQuat);
>to
>    vpQuat.mul(vpQuat, lrQuat);
>
>this will change the semantics from "rotate about the y-axis of the current
>view platform coordinates" to "rotate about the y-axis of the current
>world coordinates" which may be a more approprate semantics for you. Note
>that this difference only applies if the page-up/down keys have been
>pressed, if not, the y-axis of both coodrinates are the same.
>
>The changed semantics more closely match that of most computer games, in
>which the page-up/down keys change the pitch of one's head in the virtual
>environment, but do not change the angle of one's body relitive to the
>ground plane.




>>Dear developers,
>>
>>We are a viewplatform object connected to the scene graph and driven by a
>>transformation
>>generated by tracking devices to navigate in a virtual world. When we use
>>the keyboard
>>as input device, the right and left arrow are only turning the
viewplatform
>>around a vertical axis.
>>I verified this fact by displaying the 4x4 matrix feeded to the
viewplatform
>>and verifying that
>>only the 3x3 upper left components are modified.
>>
>>The problem I have is the following: I place two bulleyes in the world: 1
>>and 2. I translate the
>>viewplatform in the world so that I see 1 and 2 as the same point: at this
>>point my viewpoint
>>is aligned with the line passing by 1-2. I then ONLY rotate the
viewplatform
>>using the direction
>>keys. However, when I do that the viewpoint is clearly not doing only a
>>rotation because
>>1 and 2 are going appart from each others. This should not happen because
if
>>the
>>viewpoint is doing only a rotation then the viewpoint, 1, and 2 should
still
>>be aligned and
>>1 and 2 should then be seen at the same point. I verified this in OpenGL
>>fine.
>>
>>The only way that I can explained this behavior is if the projection
screen
>>instead of the
>>viewpoint was rotated. So do I miss something and I need to tune a
parameter
>>of the view
>>platform or is the perspective model wrong?
>>
>>Any help would be appreciated
>>
>>Thanks
>>
>>Yohan Baillot
>>

===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST".  For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".

Reply via email to