0
$\begingroup$

Consider the following system of 3 linear equations and 4 unknowns, represented in matrix multiplication form as follows:

$$\begin{bmatrix}C_1&C_2&C_3&C_4\end{bmatrix}\begin{bmatrix} \cos(t) \\ \sin(t) \\ u \\ v \end{bmatrix}=C_0$$

where $C_i$ ($0\leq i\leq 4$) is a 3D column vector.

How can this system be solved for $u$, $v$ and $t$?

Edit:

Method 1: (Ross Millikan's answer)

Let $c=\cos(t)$. We have $\sin(t)=\pm\sqrt{1-c^2}$.

Here $C_i^2=\begin{bmatrix}c_{i,1}^2&c_{i,2}^2&c_{i,3}^2\end{bmatrix}^T$

$$(1-c^2)C_2^2=(C_0-cC_1-uC_3-vC_4)^2$$

$$C_2^2-C_0^2=c^2(C_1^2+C_2^2)+u^2C_3^2+v^2C_4^2-2cC_0C_1-2uC_0C_3-2vC_0C_4+2cuC_1C_3+2cvC_1C_4+2uvC_3C_4$$

And it stops here; anybody knows how to solve this?

Method 2:

Let $A=\begin{bmatrix}C_1&C_2&C_3&C_4\end{bmatrix}$ and $X=\begin{bmatrix}\cos(t)&\sin(t)&u&v\end{bmatrix}^T$.

The system becomes:

$$AX=C_0$$

Using the pseudo-inverse of $A$:

$$X=A^T(AA^T)^{-1}C_0$$

$$A^T=\begin{bmatrix}C_1^T\\C_2^T\\C_3^T\\C_4^T\end{bmatrix}\quad and\quad (AA^T)^{-1}=(C_1C_1^T+C_2C_2^T+C_3C_3^T+C_4C_4^T)^{-1}$$

$AA^T$ and $C_iC_i^T$ are symmetric 3-by-3 matrices.

However I don't know of a way to expand $(AA^T)^{-1}$ while keeping $C_i$ intact.

$\endgroup$
  • $\begingroup$ You don't actually have "four unknowns", you have three: t, u, and v. These four equations are $C_1cos(t)= C_0$, $C_2sin(t)= C_0$, $C_2u= C_0$, and $C_4v= C_0$. The last two give $u= \frac{C_0}{C_3}$ and $v= \frac{C_0}{C_4}$, of course. To have $C_1cos(t)= C_0$ and $C_2 sin(t)= C_0$ you must have $\frac{C_2 sin(t)}{C_1 cos(t)}= 1$ or $tan(t)= \frac{C_1}{C_2}$. $\endgroup$ – user247327 Jun 9 '17 at 0:59
1
$\begingroup$

Let $c=\cos(t), \sin (t)= \pm \sqrt {1-c^2}$ and you have three equations in three unknowns. You have to solve it twice, once for each sign of $\sin(t)$ and getting rid of the square root may introduce extraneous solutions.

$\endgroup$
  • $\begingroup$ Your suggestion didn't really solve the problem. After getting rid of the square root, the equation becomes more complex to solve. $\endgroup$ – R. 久蔵 Jun 9 '17 at 14:42
0
$\begingroup$

Suppose $v=0$, the system becomes:

$$\begin{bmatrix}C_1&C_2&C_3\end{bmatrix}\begin{bmatrix}\cos(t)\\\sin(t)\\u\end{bmatrix}=C_0$$

which can be solved using Cramer's rule.

I'm still searching for the general case.

Edit:

Here's what I came up with using Gaussian elimination.

The system is:

$$\begin{cases}\cos⁡(t)c_{11}+\sin⁡(t)c_{12}+uc_{13}+vc_{14}=d_1\\\cos⁡(t) c_{21}+\sin⁡(t)c_{22}+uc_{23}+vc_{24}=d_2\\\cos⁡(t)c_{31}+\sin⁡(t)c_{32}+uc_{33}+vc_{34}=d_3\end{cases}$$

The augmented matrix is:

$$\begin{bmatrix}c_{11}&c_{12}&c_{13}&c_{14}&d_1\\c_{21}&c_{22}&c_{23}&c_{24}&d_2\\c_{31}&c_{32}&c_{33}&c_{34}&d_3\end{bmatrix}$$

Applying $c_{14}L_2-c_{24}L_1→L_2$ and $c_{14}L_3-c_{34}L_1→L_3$ gives:

$$\begin{bmatrix}c_{11}&c_{12}&c_{13}&c_{14}&d_1\\c_{14}c_{21}-c_{24}c_{11}&c_{14}c_{22}-c_{24}c_{12}&c_{14}c_{23}-c_{24}c_{13}&0&c_{14}d_2-c_{24}d_1\\c_{14}c_{31}-c_{34}c_{11}&c_{14}c_{32}-c_{34}c_{12}&c_{14}c_{33}-c_{34}c_{13}&0&c_{14}d_3-c_{34}d_1\end{bmatrix}$$

And applying $(c_{14}c_{23}-c_{24}c_{13})L_3-(c_{14}c_{33}-c_{34}c_{13})L_2→L_3$ gives:

$$\begin{bmatrix}c_{11}&c_{12}&c_{13}&c_{14}&d_1\\c_{14}c_{21}-c_{24}c_{11}&c_{14}c_{22}-c_{24}c_{12}&c_{14}c_{23}-c_{24}c_{13}&0&c_{14}d_2-c_{24}d_1\\a&b&0&0&m\end{bmatrix}$$

where:

$$\begin{array}{l} a=(c_{14}c_{23}-c_{24}c_{13})(c_{14}c_{31}-c_{34}c_{11})-(c_{14}c_{33}-c_{34}c_{13})(c_{14}c_{21}-c_{24}c_{11}) \\ b=(c_{14}c_{23}-c_{24}c_{13})(c_{14}c_{32}-c_{34}c_{12})-(c_{14}c_{33}-c_{34}c_{13})(c_{14}c_{22}-c_{24}c_{12}) \\ m=(c_{14}c_{23}-c_{24}c_{13})(c_{14}d_3-c_{34}d_1)-(c_{14}c_{33}-c_{34}c_{13})(c_{14}d_2-c_{24}d_1) \end{array}$$

At this point: $a\cos⁡(t)+b\sin⁡(t)=m$

And solving for $t$ gives:

$$\begin{array}{l} t≡\arctan(am\pm b\sqrt{a^2+b^2-m^2}, bm\mp a\sqrt{a^2+b^2-m^2}) [2π] \\ v={c_{14}d_2-c_{24}d_1-(c_{14}c_{21}-c_{24}c_{11})\cos⁡(t)-(c_{14}c_{22}-c_{24}c_{12})\sin⁡(t) \over c_{14}c_{23}-c_{24}c_{13}} \\ u={d_1-c_{11}\cos⁡(t)-c_{12}\sin⁡(t)-c_{13}v \over c_{14}} \end{array}$$

However, this is under these conditions:

$$c_{14}\neq0,\space c_{14}c_{23}\neq c_{24}c_{13}\space and\space \begin{cases}am\pm b\sqrt{a^2+b^2-m^2}\neq0\\bm\mp a\sqrt{a^2+b^2-m^2}\neq0\end{cases}$$

Still not optimal...

$\endgroup$

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.