Considering my old log file further, it looks like the hitch is definitely
caused by the sudden switch from (yaw=0, roll=0) to (yaw=+/-180,
roll=+/-180).
89.787949 0.000000 0.000000
89.796173 0.000000 0.000000
89.950935 -0.000000 0.000000
89.989258 -0.000000 0.000000
89.796158 180.000000 180.000000
89.639793 -180.000000 -180.000000
89.543274 180.000000 180.000000
89.290382 -180.000000 -180.000000
The QuaternionAngles() function will always produce QAngles with pitch
clamped between +/-90, therefore the yaw/roll values of +/-180 will always
occur as shown in the log above.
Example:
Pitch = 91; Yaw = 0; Roll = 0;
is equal to
Pitch = 89; Yaw = 180; Roll = 180;
These angles represent the point just past pitching straight down, when the
player is starting to flip upside down.
It would then seem that to fix the hitch, I would have to either move the
entire engine away from using Euler Angles, or I would have to change the
motion blur system. These changes aren't really in the scope of my humble
mod. For now I will settle with turning off motion blur.
Unfortunately, I have a more important gameplay bug that has been lurking
from the beginning. I haven't mentioned it until now. When I pitch
downwards towards the ground very, very slowly, the player stops pitching
right around pitch = 89, straight down. So just as the player is about to
look straight down, the slow downward pitch gets stuck (the bug also occurs
when looking straight up). This is what it looks like in the log:
Pitch Yaw Roll look_angle (this is the amount of
pitch being applied in a given frame)
89.964256 0.000000 0.000000 -0.049802
89.981636 0.000000 0.000000 -0.017369
89.985939 0.000000 0.000000 -0.032433
89.964264 0.000000 0.000000 -0.049802
89.969955 0.000000 0.000000 -0.005686
89.975136 0.000000 0.000000 -0.054909
89.981659 0.000000 0.000000 -0.043205
89.999550 0.000000 0.000000 -0.018781
So for very small pitch increments, the player's rotation gets stuck when
the pitch nears 90 degrees (straight down). You can see the pitch values
fluctuating up and down over these several frames. Note, the pitch doesn't
get stuck with larger pitch increments, so it definitely has to do with the
small look_angle values.
My initial theory was that there was some lingering pitch clamping
somewhere in the code, but I believe I have removed all instances of
clamping (Tony's article made a point of removing pitch clamping from
various places in the code).
If anyone has any advice, I'd love to hear it.
I also wanted to thank Saul, Minh, Joel and Tony for getting me this far.
You guys are awesome and I am really humbled by your generous advice.
Thanks!
On Sat, Mar 3, 2012 at 11:23 PM, Joel R. <joelru...@gmail.com> wrote:
> I think I ran into a similar problem building the orientation state using
> AxisAngles. I believe you have to first build the "default" state before
> you can multiply the (yaw/pitch/roll) quaternions.
>
>
> On Sun, Mar 4, 2012 at 12:31 AM, Craig Louie <craiglo...@gmail.com> wrote:
>
>> I've tried both Joel and Tony's methods, but I'm still unsure of how I
>> can escape the inherent aliasing problem of Euler Angles.
>>
>> With Joel's method, for instance, after I have used Slerp to get a
>> rotation quaternion, I still need to convert the resulting quaternion back
>> into Euler Angles (with QuaternionAngles()) because the Source engine
>> ultimately needs orientation in QAngle form when I call
>> engine->SetViewAngles(viewangles). Any thoughts on this?
>>
>> Also, I've learned more about the nature of the twisting hitch
>> experienced when looking straight up/down (this was my original problem).
>> When I turn off motion blur, the hitch isn't visible. My theory is that the
>> twisting hitch I am seeing is a result of the player's orientation twisting
>> at the instant the player looks straight up/down.
>>
>> Below is a part of a log showing the changes in viewangle as I pitch down
>> through pitch=90 (as shown in my youtube video).
>>
>> 88.025864 0.000000 0.000000
>> 88.136208 0.000000 0.000000
>> 88.278770 0.000000 0.000000
>> 88.421890 0.000000 0.000000
>> 88.531670 0.000000 0.000000
>> 88.683762 0.000000 0.000000
>> 88.784576 0.000000 0.000000
>> 88.976646 0.000000 0.000000
>> 89.037476 0.000000 0.000000
>> 89.261894 0.000000 0.000000
>> 89.290375 0.000000 0.000000
>> 89.522987 0.000000 0.000000
>> 89.543274 0.000000 0.000000
>> 89.787949 0.000000 0.000000
>> 89.796173 0.000000 0.000000
>> 89.950935 -0.000000 0.000000
>> 89.989258 -0.000000 0.000000
>> 89.796158 180.000000 180.000000
>> 89.639793 -180.000000 -180.000000
>> 89.543274 180.000000 180.000000
>> 89.290382 -180.000000 -180.000000
>> 89.266182 180.000000 180.000000
>> 89.037491 -180.000000 -180.000000
>> 88.784592 180.000000 180.000000
>> 88.531693 -180.000000 -180.000000
>> 88.334381 180.000000 180.000000
>> 88.278809 -180.000000 -180.000000
>> 88.025909 180.000000 180.000000
>>
>> The sudden yaw and roll change to +/-180 is probably what is causing the
>> hitch. The motion blur is creating a blur effect as if the player is
>> spinning for an instant. How to get around this, I am not sure.
>>
>>
>> Here's my current code in in_joystick.cpp's JoyStickMove() function:
>>
>> //###############################################################################
>>
>> Vector vec_for, vec_rt, vec_up, vec_for_out, vec_rt_out, vec_up_out;
>> Quaternion viewangles_q, turn_q, look_q, roll_q, qTemp, qDelta,
>> qResult, viewangles_q_out;
>>
>> AngleVectors(viewangles, &vec_for, &vec_rt, &vec_up);
>>
>> //Converting the current viewangles to a Quaternion (viewangles_q).
>> RadianEuler viewangles_euler(viewangles);
>> AngleQuaternion(viewangles_euler, viewangles_q);
>>
>> //Build a set of quaternion transforms that rotate around
>> corresponding axes in local space by corresponding angles.
>> // NOTE: non-unit vectors will bias the resulting rotation angle (but
>> not the axis)
>>
>> VectorNormalize(vec_up);
>> AxisAngleQuaternion(vec_up, turn_angle, turn_q);
>> VectorNormalize(vec_rt);
>> AxisAngleQuaternion(vec_rt, look_angle, look_q);
>> VectorNormalize(vec_for);
>> AxisAngleQuaternion(vec_for, roll_angle, roll_q);
>>
>> //Combine Quaternions.
>> QuaternionMult(look_q, turn_q, qTemp);
>> QuaternionMult(qTemp, roll_q, qDelta);
>>
>> //Slerp
>> QuaternionSlerp(viewangles_q, qDelta, 1.0, qResult);
>>
>> //Applying the Quaternion rotation.
>> QuaternionMult(qResult, viewangles_q, viewangles_q_out);
>>
>> //Converting Quaternion into a QAngle for
>> engine->SetViewAngles(viewangles_out)
>> QAngle viewangles_out;
>> QuaternionAngles(viewangles_q_out, viewangles_out);
>>
>> COM_Log("viewangle_hitch.txt", "%2f %2f %2f \n", viewangles_out.x,
>> viewangles_out.y, viewangles_out.z);
>>
>> if (viewangles_out[PITCH] > 180)
>> {
>> viewangles_out[PITCH] -= 360;
>> }
>> else if (viewangles_out[PITCH] < -180)
>> {
>> viewangles_out[PITCH] += 360;
>> }
>>
>> if (viewangles_out[YAW] > 180)
>> {
>> viewangles_out[YAW] -= 360;
>> }
>> else if (viewangles_out[YAW] < -180)
>> {
>> viewangles_out[YAW] += 360;
>> }
>>
>> if (viewangles_out[ROLL] > 180)
>> {
>> viewangles_out[ROLL] -= 360;
>> }
>> else if (viewangles_out[ROLL] < -180)
>> {
>> viewangles_out[ROLL] += 360;
>> }
>>
>> //###############################################################################
>>
>>
>> Thanks!
>>
>>
>>
>> On Fri, Mar 2, 2012 at 12:40 AM, Craig Louie <craiglo...@gmail.com>wrote:
>>
>>> Thanks Tony and Joel for more awesome help. I didn't get as much time as
>>> I would have liked to work on it tonight. I will have to continue tomorrow
>>> night (and probably into the weekend). I'll keep the thread posted. Thanks!
>>>
>>> P.S. cool flight sim vid, Joel. Are you a fellow X-Wing/Tie Fighter fan?
>>> Freespace perhaps?
>>>
>>>
>>>
>>> On Thu, Mar 1, 2012 at 10:29 PM, Joel R. <joelru...@gmail.com> wrote:
>>>
>>>> http://www.youtube.com/watch?v=5y6a_oFtNiI
>>>>
>>>>
>>>> On Thu, Mar 1, 2012 at 10:35 PM, Tony "omega" Sergi <
>>>> omegal...@gmail.com> wrote:
>>>>
>>>>> This actually will work in Source, as the input/view stuff is
>>>>> basically identical:
>>>>> http://articles.thewavelength.net/269/
>>>>> -Tony
>>>>>
>>>>>
>>>>>
>>>>> On Fri, Mar 2, 2012 at 5:09 AM, Craig Louie <craiglo...@gmail.com>wrote:
>>>>>
>>>>>> Thanks Joel! I'll also try out your method when I get home. I'll keep
>>>>>> everyone posted on my results.
>>>>>>
>>>>>> Thanks!
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Thu, Mar 1, 2012 at 11:12 AM, Joel R. <joelru...@gmail.com> wrote:
>>>>>>
>>>>>>> Quaternion.h
>>>>>>> http://pastebin.com/hxTBab3T
>>>>>>>
>>>>>>> Quaternion.cpp
>>>>>>> http://pastebin.com/2WizAewD
>>>>>>>
>>>>>>> Quaternion Rotation Formulas (Yaw/Pitch/Roll)
>>>>>>> http://pastebin.com/BHRvcbCG
>>>>>>>
>>>>>>> I got most of the quaternion stuff from Ogre engine and converted to
>>>>>>> work on Source Engine. The functions you will be interested in using
>>>>>>> are
>>>>>>> FromRotationMatrix, ToRotationMatrix, AxisAngle, ToAxisAngle,
>>>>>>> GetRotationTo, ToAxes, Slerp, GetAngles, GetMatrix
>>>>>>>
>>>>>>> So basically you'll do what's below: (this is psuedocode, function
>>>>>>> names are not exact!)
>>>>>>>
>>>>>>>
>>>>>>> Quat current, target;
>>>>>>> current.FromRotationMatrix( ent->EntityWorldTransform() );
>>>>>>>
>>>>>>> target = current;
>>>>>>> target = yaw( target, 10 ); //10 degrees to the right, this function
>>>>>>> is not part of Quat class for some reason
>>>>>>>
>>>>>>> current = Slerp( current, target, 0.2 ); //This is only 20%, you'll
>>>>>>> need to calculate the angle difference between current/target, and then
>>>>>>> move based on your angles per second speed.
>>>>>>>
>>>>>>> matrix3x4_t newRot;
>>>>>>> current.ToRotationMatrix( newRot );
>>>>>>> ent->SetWorldTransform( newRot );
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Enjoy!
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On Thu, Mar 1, 2012 at 12:46 PM, Joel R. <joelru...@gmail.com>wrote:
>>>>>>>
>>>>>>>> The quaternion calculations you're doing are similar to just
>>>>>>>> manipulating the Euler Angles, which cause Gimbal Lock.
>>>>>>>>
>>>>>>>> The way I did it in my flight simulation is to create a Current
>>>>>>>> Quaternion (from the starting Matrix Transformation when user spawns)
>>>>>>>> and
>>>>>>>> Target Quaternion (where they want to rotate to). I also used some
>>>>>>>> formulas to rotate the quaternion in the individual yaw/pitch/roll
>>>>>>>> orientations, I recommend googling for these formulas. I then used the
>>>>>>>> Slerp formula to animate from the Current quaternion to the Target
>>>>>>>> quaternion. Slerp will avoid Gimbal-Lock entirely, which is very neat.
>>>>>>>>
>>>>>>>> After you do those calculations, just convert the new Current
>>>>>>>> Quaternion into a Matrix Transformation and set the Entity to take that
>>>>>>>> transform.
>>>>>>>>
>>>>>>>> I've used this to control the entity rotation as well as the
>>>>>>>> camera, it allows you to do some pretty nifty tricks.
>>>>>>>>
>>>>>>>> I'll look for the code, if I still have it, and I'll post it here
>>>>>>>> for you. Or if you figure out, please post a video =D
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>>
>>>>>>>> Joel
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On Thu, Mar 1, 2012 at 11:55 AM, Minh <minh...@telus.net> wrote:
>>>>>>>>
>>>>>>>>> You can try spitting out the results to a file so you can observe
>>>>>>>>> the hitch more closely.
>>>>>>>>> Look for a function called COM_Log
>>>>>>>>>
>>>>>>>>> COM_Log ( "viewangle_hitch.txt", "%.2f %.2f %.2f \n",
>>>>>>>>> viewangles.x, viewangles.y, viewangles.z );
>>>>>>>>>
>>>>>>>>> On 3/1/2012 3:55 AM, Craig Louie wrote:
>>>>>>>>>
>>>>>>>>> Thanks for the reply Saul. I hadn't thought of disabling
>>>>>>>>> prediction. I just gave it a shot. I changed to "engine->ClientCmd(
>>>>>>>>> "cl_predict 0" );" in cdll_client_int.cpp. With another test, the
>>>>>>>>> unsmooth
>>>>>>>>> rotation when looking straight up/down still occurs.
>>>>>>>>>
>>>>>>>>> My main theories at this point:
>>>>>>>>> 1) Something is occurring in VectorAngles(vec_for_out, vec_up_out,
>>>>>>>>> viewangles) to cause this. VectorAngles() might not produce straight
>>>>>>>>> up/down Euler angles properly - choking because of aliasing perhaps?
>>>>>>>>> This
>>>>>>>>> is a big guess because I don't know the inner workings of
>>>>>>>>> VectorAngles().
>>>>>>>>> 2) There is some code somewhere else that is conflicting with the
>>>>>>>>> full, unclamped rotation I've implemented. I've tried to remove all
>>>>>>>>> instances of pitch clamping, for example. But I may have missed
>>>>>>>>> something.
>>>>>>>>> Also, it's possible mouse rotation is somehow conflicting with my
>>>>>>>>> changes
>>>>>>>>> to joystick rotation.
>>>>>>>>>
>>>>>>>>> Here's a video of the issue I'm seeing. It happens a few times in
>>>>>>>>> the video. You'll see it happen the first time at about 11 seconds in
>>>>>>>>> when
>>>>>>>>> the text on the floor twists/shakes.
>>>>>>>>> http://www.youtube.com/watch?v=PLN641Qsn08
>>>>>>>>>
>>>>>>>>> Thanks!
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Wed, Feb 29, 2012 at 11:34 PM, Saul Rennison <
>>>>>>>>> saul.renni...@gmail.com> wrote:
>>>>>>>>>
>>>>>>>>>> The hitching is most likely something trying to revert your angle
>>>>>>>>>> changes (I.e. in prediction). Have you tried disabling prediction?
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On Thursday, March 1, 2012, Craig Louie wrote:
>>>>>>>>>>
>>>>>>>>>>> Hello Coders,
>>>>>>>>>>>
>>>>>>>>>>> I'm trying to create a space flight combat game with the Source
>>>>>>>>>>> SDK. Currently, I am trying to achieve rotation around the player's
>>>>>>>>>>> local
>>>>>>>>>>> axes for airplane-like pitch, yaw and roll. Below is the code I'm
>>>>>>>>>>> using to
>>>>>>>>>>> try to implement full 3D rotation. I'm running into a problem where
>>>>>>>>>>> the
>>>>>>>>>>> player's view twists and hitches unsmoothly when the player looks
>>>>>>>>>>> straight
>>>>>>>>>>> up or straight down. I think this may be because of the inherent
>>>>>>>>>>> problem of
>>>>>>>>>>> aliasing/Gimbal-lock when using Euler angles. Any advise?
>>>>>>>>>>>
>>>>>>>>>>> //#########################################
>>>>>>>>>>> //Take the Euler angles for the player's orientation
>>>>>>>>>>> (viewangles) and convert them to vectors.
>>>>>>>>>>> Vector vec_for, vec_rt, vec_up, vec_for_out, vec_rt_out,
>>>>>>>>>>> vec_up_out;
>>>>>>>>>>> AngleVectors(viewangles, &vec_for, &vec_rt, &vec_up);
>>>>>>>>>>>
>>>>>>>>>>> //Take the resulting vectors and the desired rotation amounts
>>>>>>>>>>> and convert them to Quaternions.
>>>>>>>>>>> Quaternion turn_q, look_q, roll_q;
>>>>>>>>>>> // NOTE: Assumes axis is a unit vector, non-unit vectors will
>>>>>>>>>>> bias the resulting rotation angle (but not the axis)
>>>>>>>>>>> VectorNormalize(vec_up);
>>>>>>>>>>> AxisAngleQuaternion(vec_up, turn_angle, turn_q);
>>>>>>>>>>>
>>>>>>>>>>> VectorNormalize(vec_rt);
>>>>>>>>>>> AxisAngleQuaternion(vec_rt, look_angle, look_q);
>>>>>>>>>>>
>>>>>>>>>>> VectorNormalize(vec_for);
>>>>>>>>>>> AxisAngleQuaternion(vec_for, roll_angle, roll_q);
>>>>>>>>>>>
>>>>>>>>>>> //Convert the quaternions to matrices.
>>>>>>>>>>> matrix3x4_t xform0, xform1, xform2;
>>>>>>>>>>> QuaternionMatrix(turn_q, xform0);
>>>>>>>>>>> QuaternionMatrix(look_q, xform1);
>>>>>>>>>>> QuaternionMatrix(roll_q, xform2);
>>>>>>>>>>>
>>>>>>>>>>> //Combine matrices.
>>>>>>>>>>> matrix3x4_t xformA, xformB;
>>>>>>>>>>> MatrixMultiply(xform0, xform1, xformA);
>>>>>>>>>>> MatrixMultiply(xformA, xform2, xformB);
>>>>>>>>>>>
>>>>>>>>>>> //Rotate the vectors with our combined rotation matrix.
>>>>>>>>>>> VectorRotate(vec_for, xformB, vec_for_out);
>>>>>>>>>>> VectorRotate(vec_rt, xformB, vec_rt_out);
>>>>>>>>>>> VectorRotate(vec_up, xformB, vec_up_out);
>>>>>>>>>>>
>>>>>>>>>>> //Determine the new viewangles based on our rotated vectors.
>>>>>>>>>>> VectorAngles(vec_for_out, vec_up_out, viewangles);
>>>>>>>>>>> //#########################################
>>>>>>>>>>>
>>>>>>>>>>> I put this code into in_joystick.cpp in the JoyStickMove( float
>>>>>>>>>>> frametime, CUserCmd *cmd ) function.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Thanks!
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> --
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Kind regards,
>>>>>>>>>> *Saul Rennison*
>>>>>>>>>>
>>>>>>>>>> _______________________________________________
>>>>>>>>>> To unsubscribe, edit your list preferences, or view the list
>>>>>>>>>> archives, please visit:
>>>>>>>>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> To unsubscribe, edit your list preferences, or view the list
>>>>>>>>> archives, please
>>>>>>>>> visit:https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> To unsubscribe, edit your list preferences, or view the list
>>>>>>>>> archives, please visit:
>>>>>>>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> To unsubscribe, edit your list preferences, or view the list
>>>>>>> archives, please visit:
>>>>>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> To unsubscribe, edit your list preferences, or view the list
>>>>>> archives, please visit:
>>>>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> -Tony
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> To unsubscribe, edit your list preferences, or view the list archives,
>>>>> please visit:
>>>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>>
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> To unsubscribe, edit your list preferences, or view the list archives,
>>>> please visit:
>>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>>>
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> To unsubscribe, edit your list preferences, or view the list archives,
>> please visit:
>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>>
>>
>>
>
> _______________________________________________
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
>
>
>
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please
visit:
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlcoders
