4
$\begingroup$

Context

I am writing a "flight" trajectory program in Mathematica that ignores all physics. I model a turning aircraft (pitching at full deflection) as a parametrised circle. I have been able to simulate instantaneous rolls of the aircraft by applying a rotation at an instant in time. I would now like to simulate a continuous roll of the aircraft; that is, applying a rotation at a constant rate over an interval of time as the aircraft continues to move along its turn circle (pitching at full deflection). In the following section, I provide examples of two instantaneous rolls.

The problem

I have a circle $C$ parametrised as $$\begin{pmatrix}\cos(t) \\ \sin(t)\\0\end{pmatrix}$$ and a transformation $R_{t_1,\theta}$ that rotates a parametrised circle around its tangent vector at a particular time $t_1$ by an angle of $\theta$ using the Rodrigues rotation formula (Matrix for rotation around a vector).

Let $M(v,\alpha)$ be the Rodrigues rotation matrix around a vector $v$ by an angle of $\alpha$.

Then

$$R_{t_1,\theta} = M\bigg(\frac{C'(t_1)}{\lVert C'(t_1)\rVert},\theta\bigg)(C-C(t_1))+C(t_1)$$

For example, let $C_{1} = R_{\frac{\pi}{8},\frac{\pi}{16}}(C)$. The plotted result is shown below with $C$ in blue and $C_1$ in yellow. $C(\frac{\pi}{8})$ is shown in red:

enter image description here

We can apply the procedure again to obtain $C_2 = R_{\frac{\pi}{2},\frac{\pi}{16}}(C_1)$: enter image description here

Is there a "continuous" rotation $R^c$ that rotates a circle $C$ at a rate of $\theta$ per unit of time over an interval $[t_1,t_2]$?

i.e. $R^c_{t_1,t_2,\theta}(C)=\lim_{N\to\infty}R_{t_1+\frac{N(t_2-t_1)}{N},\frac{\theta}{t_2-t_1}/N}\bigg(R_{t_1+\frac{(N-1)(t_2-t_1)}{N},\frac{\theta}{t_2-t_1}/N}\bigg(...R_{t_1+\frac{t_2-t_1}{N},\frac{\theta}{t_2-t_1}/N}(C)...\bigg)$

Here is the Mathematica code used to generate the plots:

f[v_] := {{0, -v[[3]], v[[2]]}, {v[[3]], 0, -v[[1]]}, {-v[[2]], 
   v[[1]], 0}}

getMat[v_, p_] := {{1, 0, 0}, {0, 1, 0}, {0, 0, 1}} + 
  Sin[p]*f[v] + (2*Sin[p/2]^2)*f[v].f[v]

rotate[time_, amount_, c_] :=
    Module[{p = c[time],
           d = Normalize[D[c[t], t] /. t -> time]},
    Return[getMat[d, amount].(c[t] - p) + p]
  ]

c[t_] := {Cos[t], Sin[t], 0}

c1 = Function[t, Evaluate[Simplify[rotate[Pi/8, Pi/16, c]]]];

c2 = Function[t, Evaluate[Simplify[rotate[Pi/2, Pi/16, c1]]]];

Show[
  ParametricPlot3D[{c[u], c1[u], c2[u]}, {u, 0,
   2 \[Pi]}, AxesLabel -> {xa, y, z}] ,
 Graphics3D[{PointSize[0.02], Red, Point[c[Pi/8]]}],
 Graphics3D[{PointSize[0.02], Blue, Point[c1[Pi/2]]}]
 ]

Update:

After experimenting with a simulation and plotting the points of rotation up to $R_{\frac{13\pi}{8},\frac{\pi}{64}}(R_{\frac{12\pi}{8},\frac{\pi}{64}}(R_{\frac{11\pi}{8},\frac{\pi}{64}}(...R_{0,\frac{\pi}{64}}(C)...) $

enter image description here

I conjecture that the plot of ${R^c}_{0,t,\theta}(C)(t), t > 0$ is a spiral whose axis of rotation is tilted by some amount:

enter image description here

If so, finding the expression for this spiral should be sufficient to derive $R^c$. I think in order to find the spiral, I need an expression that relates the turn (pitch) rate (constant 1 radian per unit of time) and rotation (roll) rate ($\theta$ per unit of time) to the axis of the spiral.

$\endgroup$
  • 1
    $\begingroup$ Since you're using Mathematica: I gave an implementation of the Rodrigues rotation formula here. $\endgroup$ – J. M. is a poor mathematician Mar 19 '18 at 17:47
  • 2
    $\begingroup$ You might find infinitesimal rotations useful. See, for instance, web.uconn.edu/~ch351vc/pdfs/infrot.pdf, astro.caltech.edu/~golwala/ph125ab/ph106ab_notes_sec5.1.pdf or rotations.berkeley.edu/infinitesimal-rotations $\endgroup$ – amd Mar 20 '18 at 1:20
  • $\begingroup$ @J.M.isnotamathematician Cheers! Also, thanks for the answer I linked on the Rodrigues formula! $\endgroup$ – ignoramus Mar 20 '18 at 7:47
  • $\begingroup$ @amd Thanks I will check it out! $\endgroup$ – ignoramus Mar 20 '18 at 7:47
  • 1
    $\begingroup$ There’s no need to code up Rodrigues’ formula for yourself. Mathematica has a very flexible built-in function RotationTransform. In fact, you could replace the entire rotate cascade with something like RotationTransform[amount,D[c[t], t] /. t -> time,c[time]]@c[t]. $\endgroup$ – amd Mar 20 '18 at 19:12
0
$\begingroup$

After applying the infinitesimal rotations suggested by amd, I have derived an expression for ${R^c}_{t_1,t_2,\theta}(C)(t)$.

First, we represent the velocity of the aircraft over the period of continuous rotation as $\mathbf{v}(t)$. The pitch axis $\mathbf{c}(t)$ is orthogonal to the velocity and lift vectors, and we choose it to be pointing right initially, relative to the forward motion of the aircraft.

Let $R(\mathbf{x,\phi})$ be the infinitesimal Rodrigues rotation matrix around the unit vector $\mathbf{x}$ by an angle of $\phi$. Then $\mathbf{c}$ is being rotated at a rate of $\theta$ (rolling) per unit of time around $\mathbf{v}$, while $\mathbf{v}$ is being rotated at a rate of $\alpha$ (pitching) per unit of time around $\mathbf{c}$.

Then

$$ d\mathbf{v} = R(\mathbf{c},\alpha)\mathbf{v}-\mathbf{v} = (R(\mathbf{c},\alpha)-I)\mathbf{v}$$ $$ d\mathbf{c} = R(\mathbf{v},\theta)\mathbf{c}-\mathbf{c} = (R(\mathbf{v},\theta)-I)\mathbf{c}.$$

After simplifying, we obtain

$$ \frac{d\mathbf{v}}{dt} = \alpha \mathbf{c} \times \mathbf{v}$$ $$ \frac{d\mathbf{c}}{dt} = \theta \mathbf{v} \times \mathbf{c},$$

which implies $$ \frac{d\mathbf{v}}{dt} = -\frac{\alpha}{\theta} \frac{d\mathbf{c}}{dt}$$ $$ \frac{d\mathbf{c}}{dt} = -\frac{\theta}{\alpha} \frac{d\mathbf{v}}{dt}.$$

Integrating both sides w.r.t. $t$, we have $$ \mathbf{v} = -\frac{\alpha}{\theta} \mathbf{c}+\mathbf{k_0}$$ $$ \mathbf{c} = -\frac{\theta}{\alpha} \mathbf{v}+\mathbf{k_1}$$ $$ \mathbf{k_0} = \mathbf{v}(0)+\frac{\alpha}{\theta} \mathbf{c}(0)$$ $$ \mathbf{k_1} = \mathbf{c}(0)+\frac{\theta}{\alpha} \mathbf{v}(0).$$

Substituting $\mathbf{k}_0$ and $\mathbf{k}_1$ back into the differential equations, we obtain

$$\frac{d\mathbf{v}}{dt} = (\theta \mathbf{v}(0) + \alpha \mathbf{c}(0))\times \mathbf{v} $$ $$\frac{d\mathbf{c}}{dt} = (\theta \mathbf{v}(0) + \alpha \mathbf{c}(0))\times \mathbf{c} .$$

These equations have solutions as given in this answer here:

Differential equation involving cross product.

We then obtain the translation of $C$ at $t_2$ by integrating $\mathbf{v}$.

Let

$$M = \frac{C(0)+C(\pi)}{2}$$ $$\mathbf{f} = \int \mathbf{v} dt$$ $$\mathbf{f}(0) = C(t_1)$$ $$\mathbf{v}(0) = C'(t_1)$$ $$\mathbf{c}(0) = (C(t_1)-M)\times \mathbf{v}(0).$$

Then the rotated circle is a circle that tangentially intersects the spiral $\mathbf{f}$, with a time offset of $t_2$. In addition, since $C$ is the unit circle with period $2\pi$, $\alpha=1$.

The final expression is then

$${R^c}_{t_1,t_2,\theta}(C)(t) = \mathbf{f}(t_2-t_1)+\mathbf{c}(0)\times\mathbf{v}(0)+\cos{(t-t_2)}\mathbf{v}(t_2-t_1)\times\mathbf{c}(t_2-t_1)+\sin{(t-t_2)}\mathbf{v}(t_2-t_1).$$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.