[PD] gimbal lock solution?
Iam trying to rotate GEM object with my android phone, sending from acc sensor througth OSC Gem object rotateXYZ is not the solution for that, maybe because of gimbal lock effect... Did anybody similar working patch in pd ? Or can somebody hel me ? many thanks. fk. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
On Sam, 2013-02-09 at 13:07 +0100, Fero Kiraly wrote: Iam trying to rotate GEM object with my android phone, sending from acc sensor througth OSC Gem object rotateXYZ is not the solution for that, maybe because of gimbal lock effect... Did anybody similar working patch in pd ? Or can somebody hel me ? many thanks. If I understand the gimbal lock correctly (I only read about it now), your problem seems not related to it. With accelerometer data only, you can only calculate two axis of orientation (pitch and roll), while yaw is still missing. You cannot measure the angle of rotation along the direction of gravity with accelerometers. You'd need something else for yaw, a compass or a gyroscope (while the latter is giving you only relative measurements). Roman ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
I have 3 axis accelerometer and 3 axis orientation sensor in my device. So i can get 0 - 360 deg on X, Y and Z. When I draw in GEM a rectangle, I am able to rotate it with each of 3 axes, but only separately. When I try it with all three axes together it has strange movements, so I guess it has something with gimbal lock. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
hello, rotateXYZ is doing : rotation X then rotation Y then rotation Z this is not the same than : rotation Z rotation Y rotation X So, depending on how works your orientation sensors, you could have to change the rotation order. you can try, using 3 rotateXYZ object. cheers c Le 09/02/2013 14:45, Fero Kiraly a écrit : I have 3 axis accelerometer and 3 axis orientation sensor in my device. So i can get 0 - 360 deg on X, Y and Z. When I draw in GEM a rectangle, I am able to rotate it with each of 3 axes, but only separately. When I try it with all three axes together it has strange movements, so I guess it has something with gimbal lock. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
I believe it is (unfortunately) not as simple as that. When you rotate with [rotateXYZ], the three axis are not always perpendicular to each other. For instance, if you set Y to 90, the X and the Z rotate around the same axis. However, the gyroscope (at least the one from Wiimote MotionPlus and most likely also the one from your phone) consists of three separate sensor which are perpendicular to each other at any time. If I understand Fero correctly, he would like to use the sensors so he can control an object in Gem that follows exactly the orientation of his phone. I also tried that once with a Wiimote and Gem but I couldn't wrap my head around the two concepts of [rotateXYZ] and the gyro sensors. Anyone enlightened might be able to shed some light on this confusion? Roman On Sam, 2013-02-09 at 15:18 +0100, Cyrille Henry wrote: hello, rotateXYZ is doing : rotation X then rotation Y then rotation Z this is not the same than : rotation Z rotation Y rotation X So, depending on how works your orientation sensors, you could have to change the rotation order. you can try, using 3 rotateXYZ object. cheers c Le 09/02/2013 14:45, Fero Kiraly a écrit : I have 3 axis accelerometer and 3 axis orientation sensor in my device. So i can get 0 - 360 deg on X, Y and Z. When I draw in GEM a rectangle, I am able to rotate it with each of 3 axes, but only separately. When I try it with all three axes together it has strange movements, so I guess it has something with gimbal lock. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
Roman, thank you, that is exactly what I mean. Google said to me that it has something with: kalman or complementary filter, gimbal lock, quaternions arithmetic with matrixes ( there I can use iemmatrix library) my input data are: accX, accY, accZ, pitch yaw, roll (from HTC desire) how to get this together ? fk. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
Le 09/02/2013 15:49, Fero Kiraly a écrit : Roman, thank you, that is exactly what I mean. Google said to me that it has something with: kalman or complementary filter, gimbal lock, quaternions arithmetic with matrixes ( there I can use iemmatrix library) translate/rotate/scale/shear are matrix operation! my input data are: accX, accY, accZ, pitch yaw, roll (from HTC desire) how to get this together ? use pitch, yaw, roll in 3 rotateXYZ and connect the 3 rotate object in the correct order. (not X, Y, Z, since it did not work) cheers c fk. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
Le 09/02/2013 15:39, Roman Haefeli a écrit : I believe it is (unfortunately) not as simple as that. When you rotate with [rotateXYZ], the three axis are not always perpendicular to each other. For instance, if you set Y to 90, the X and the Z rotate around the same axis. every rotation keep the 3 axes perpendicular. but rotateXYZ do 3 rotations, X, Y and Z, in this order. if you rotate 90° in Y, Z axe became what use to be the X axe. so rotation in Z after a 90° rotation in Y is the same than a rotation in X before the Y rotation. that's the gimbal lock. but after the Y rotation, a X rotation is not the same than a Z rotation. if you don't believe me, you can try : gemhead rotate X 0 Z rotate 0 90 0 cube X and Z are still (and will always be) perpendicular. However, the gyroscope (at least the one from Wiimote MotionPlus and most likely also the one from your phone) consists of three separate sensor which are perpendicular to each other at any time. If I understand Fero correctly, he would like to use the sensors so he can control an object in Gem that follows exactly the orientation of his phone. I also tried that once with a Wiimote and Gem but I couldn't wrap my head around the two concepts of [rotateXYZ] and the gyro sensors. gyro sensors give rotation speed. not absolute rotation. in theory (with infinite acurate sensors, no drift etc), you have to do a feedback loop (using gemlist_info), so that the current rotation is modify by the sensors value. (rotation did not sums up) cheers c Anyone enlightened might be able to shed some light on this confusion? Roman On Sam, 2013-02-09 at 15:18 +0100, Cyrille Henry wrote: hello, rotateXYZ is doing : rotation X then rotation Y then rotation Z this is not the same than : rotation Z rotation Y rotation X So, depending on how works your orientation sensors, you could have to change the rotation order. you can try, using 3 rotateXYZ object. cheers c Le 09/02/2013 14:45, Fero Kiraly a écrit : I have 3 axis accelerometer and 3 axis orientation sensor in my device. So i can get 0 - 360 deg on X, Y and Z. When I draw in GEM a rectangle, I am able to rotate it with each of 3 axes, but only separately. When I try it with all three axes together it has strange movements, so I guess it has something with gimbal lock. ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] gimbal lock solution?
On Sam, 2013-02-09 at 16:41 +0100, Cyrille Henry wrote: Le 09/02/2013 15:39, Roman Haefeli a écrit : I believe it is (unfortunately) not as simple as that. When you rotate with [rotateXYZ], the three axis are not always perpendicular to each other. For instance, if you set Y to 90, the X and the Z rotate around the same axis. every rotation keep the 3 axes perpendicular. but rotateXYZ do 3 rotations, X, Y and Z, in this order. if you rotate 90° in Y, Z axe became what use to be the X axe. so rotation in Z after a 90° rotation in Y is the same than a rotation in X before the Y rotation. that's the gimbal lock. but after the Y rotation, a X rotation is not the same than a Z rotation. if you don't believe me, you can try : gemhead rotate X 0 Z rotate 0 90 0 cube X and Z are still (and will always be) perpendicular. Ok, this makes sense to me now. Thanks. However, the gyroscope (at least the one from Wiimote MotionPlus and most likely also the one from your phone) consists of three separate sensor which are perpendicular to each other at any time. If I understand Fero correctly, he would like to use the sensors so he can control an object in Gem that follows exactly the orientation of his phone. I also tried that once with a Wiimote and Gem but I couldn't wrap my head around the two concepts of [rotateXYZ] and the gyro sensors. gyro sensors give rotation speed. not absolute rotation. Yeah, right. But you can add the values up to get rotation (of course, it's always rotation relative to itself, which will obviously drift from reality over time). in theory (with infinite acurate sensors, no drift etc), you have to do a feedback loop (using gemlist_info), so that the current rotation is modify by the sensors value. (rotation did not sums up) ...parsing solution beep.. stack overflow (in my head)! You mean you add each increment (gyro measurement) to the rotation separately, then get the orientation with gem_list info, add the next increment to the orientation, get orientation again, add increment etc? Can you cast that into a little Gem example patch, if it is not too much trouble? Roman ___ Pd-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
