DEMO.DESIGN FAQ

следующий фpагмент (2)

- Usenet echoes (21:200/1) -------------------------- COMP.GRAPHICS.ALGORITHMS - Msg : 4 of 9 From : eyal@fir.canberra.edu.au 2:5030/144.99 11 Aug 94 23:49:00 To : All 13 Aug 94 01:47:18 Subj : (1) Re: Flight Simulator 3-D Equations -------------------------------------------------------------------------------- In <32brtf$66n@acmez.gatech.edu> gt6681c@prism.gatech.edu (Michael Steven Heyda) writes: >I have the basic equations to do pitch,roll,and yaw translations >on the object but those equations don't allow the ship to fly >(respond to controls) correctly. >E.g. when you pull back on the stick it should always pitch up >relative to your perspective looking out the cockpit, not pitch >up relative to the coordinate system. >The pitch problem I figured out but I can't lay my finger on the >solution to the yaw problem. i.e. when you yaw it should always >go around in a circle, not just yaw on the coordinate system but >relative to how you're looking out the cockpit. >Does anyone have some flight sim source they could share or just >point me the right direction. The trick here is to do the rotation in the correct order, then it works as you expect. The common order is to rotate in this order: yaw pitch roll Now, this will take you from fixed (world) coords to body coords which are then used for the display. However, if you want to update the attitude based on the controls then you will start with the body attitude matrix and apply the above rotation. If you want now to go back to world coords then you simply use the inverse (in this case the transpose is the same) matrix. If your body is rotating at the time and has inertia then you will want to use a 6-dof set of rules instead; these can be used in full, or in a simplified form where you ignore some of the precession effects. Another simple approach is to convert the WORLD Euler angles (heading, pitch roll) with the BODY angular rates (yaw-rate, pitch-rate, roll-rate) into WORLD angular rates (heading-rate, pitch-rate, roll-rate) and then simply add these to the world Euler angles (heading, pitch, roll). These formulaes are simple and will work fine as long as you keep the rotation small, which means the frame rate high enough, or you simulate each frame in multiple small steps. All angles/rates in radians. We start with knowing: world-pitch [known] world-yaw (heading) [known] world-roll [known] body-pitch-rate [from controls] body-yaw-rate [from controls] body-roll-rate [from controls] We end up with world-pitch-rate world-yaw-rate world-roll-rate which we add back to the world angles. The calcs are (note that the order IS important): world-pitch-rate = roll-rate * cos(roll) - yaw-rate * sin(roll) world-yaw-rate = yaw-rate * cos(roll) - pitch-rate * sin(roll) world-roll-rate = roll-rate - world-yaw-rate * sin(pitch) Now you add wold-pitch-rate to world-pitch etc. Do note that this last 'add' is really 'integrate' and how complex you make it depends on how much fidelity you need. Hpoe this clarifies the situation a bit. -- Regards Eyal Lebedinsky eyal@ise.canberra.edu.au

следующий фpагмент (3)|пpедыдущий фpагмент (1)

- Usenet echoes (21:200/1) -------------------------- COMP.GRAPHICS.ALGORITHMS - Msg : 117 of 165 From : UUCP 2:5030/144.99 08 Sep 94 16:50:22 To : All 11 Sep 94 05:36:58 Subj : (1) Re: Movement in 3d space? -------------------------------------------------------------------------------- Dear Terry, as far as I see it, you should use the following transformations: Let vp(x,y,z) be the view point in world coordinates. Let vp(x^,y^,z^) be the view point in viewing coordinates. Let vp(x¦,y¦,z¦) be the new view point in world coordinates. Apply first the x-rotation (pitch), then y-rot. (yaw) and finally z-rot. (roll). The transformation matrices are: | 1 0 0 | Rx = | 0 cos(pitch) sin(pitch) | | 0 -sin(pitch) cos(pitch) | | cos(yaw) 0 -sin(yaw) | Ry = | 0 1 0 | | sin(yaw) 0 cos(yaw) | | cos(roll) sin(roll) 0 | Rz = |-sin(roll) cos(roll) 0 | | 0 0 1 | The transformation is: vp(x^,y^,z^) = vp(x,y,z)*Rx*Ry*Rz After this transformation, apply the movement delta to the z^ coordinate of your view point (right handed coordinate system), since you have the viewing coordinates as follows: x^- and y^-axis from bottom left to bottom right and top left corners of your screen. z^ point right into your eyes, hence: a negative delta to get closer, a positive delta to get away Then apply a reverse transformation. This is easy, since the transformation matrices are orthogonal. Thus, inverse of B equals transposed of B (== transp. B) vp(x¦,y¦,z¦) = vp(x^,y^,z¦+delta)*(transp. Rz)*(transp. Ry)*(transp. Rx) Hope this is correct, since it's a long time ago that I dealed with that stuff. A tip: the book Fundamentals of Interactive Graphics by Foley/van Dam is very helpful. (Addison-Wesley Systems Programming Series) Sincerely,

следующий фpагмент (4)|пpедыдущий фpагмент (2)

- Usenet echoes (21:200/1) -------------------------- COMP.GRAPHICS.ALGORITHMS - Msg : 127 of 163 From : c2a192@ugrad.cs.ubc.ca 2:5030/144.99 08 Sep 94 19:36:36 To : All 11 Sep 94 05:37:04 Subj : (1) Yaw, Pitch and Roll... -------------------------------------------------------------------------------- A while ago someone was asking how to perform yaw, pitch and roll relative to virtual airplane's coordinate system, so that pulling back on the stick will make the plane go up with respect to the cockpit, not the ground. I recently re-written my transformation module to allow just that sort of thing, and more. A camera is represented by three orthonormal basis vectors, call them u, v, w. The observer looks "through" vector w. Vector v points up, relative to the cockpit. Vector v points to the right. As you can see, I use a right-handed coordinate systems, with objects receding into the negative-z distance. Now, the three camera basis vectors can be combined to form the columns of an orhogonal matrix Q. (orthogonal matrix = columns are mutually perpendicular, and are all unit length). You can apply a transformation, such as a rotation, to this matrix using straight multiplication, as in: Qnew = TQ where T is the transformation matrix. This method works relative to absolute coordinates, not relative to the camera coordinates. It turns out that if you want the transformation to affect the camera relative ot it's own coordinates, all you have to do is reverse the multiplication: Qnew = QT What this does now is project the transformation T into the column space of the camera Q. For instance, if T is a transformation that represents a pitch-up of 15 degrees, the transformation TQ Will rotate camera Q by 15 degrees around the X axis. However, the transformation QT will rotate the camera about vector v (the one that points left). In other words, it will perform the pitch operation relative to the camera's position. When you have your camera defined this way, the way you apply the camera to the scene is: for each point x in the scene, do T T x_new = Q x where Q is the transpose matrix of Q. Because Q is an orthogonal matrix, its inverse is simply the transpose, so no computing is required to get the inverse. The following is a formula for transforming a point relative to the camera's coordinates, rather than absolute coordinates: T x_new = Q T Q x_old The point x_old is first multiplied by the inverse of Q to convert it to a coordinate system that is relative to the camera. Once it is in this coordinate system, it is multipled by T to perform the transformation. Then, it is multiplied by Q to bring it back to absolute coordinates. In this way, we represent a complex transformation by decoupling it into a change of coordinates and a simpler transformation. Notice that to transform the camera itself about its own coordinates, the formula reduces as follows: T Q_new = QTQ Q T = QTI (Q Q = identity matrix) Q_new = QT which is just the formula given previously.

Всего 3 фpагмент(а/ов) |пpедыдущий фpагмент (3)


оглавление \| demo party в ex-СССР \| infused bytes e-mag \| новости от ib/news \| другие проекты \| письмо \|	win koi lat