2
$\begingroup$

We have four points: $p_1, p_2, p_3$ and $p_4$. The goal is to compute the distance between $p_1$ and $p_4$.

What we know:

  • The distances between $p1$ and $p2$ ($a$) , $p2$ and $p3$ ($b$), $p3$ and $p4$ ($c$)
  • The angles $\alpha$ and $\beta$ between ($p1$,$p2$,$p3$) and ($p2$,$p3$,$p4$) respectively
  • The dihedral (torsion) angle $\phi$ between the two planes of $p1$ and $p4$

We do not have any coordinates of any of these points.

How do we use this information to compute the distance between $p1$ and $p4$? I tried doing something like projecting $p1$ onto the plane of $p4$, but I ended up needing information that I do not have access to.

dihedral angle

$\endgroup$

2 Answers 2

1
$\begingroup$

In order to take into account a certain symmetry of this issue, let us consider the intersection line of the planes as the $x$ axis with origin at the midpoint of $p_2p_3$, and the bissector plane as the reference horizontal plane.

Imagine the particular case $\varphi=0$ meaning that the figure is a flat figure in this bissector plane , i.e., all $p_k$ coplanar in the $xy$ plane. From this case, generate the general case by rotating vector $\vec{p_1p_2}$ around the $x$ axis with a $\psi:=\varphi/2$ angle in one direction whereas vector $\vec{p_3p_4}$ is rotated in the other direction i.e., with a $-\psi$ angle.

This will give, using rotation matrices,

$$p_1=\begin{pmatrix}1&0&0\\0&\cos \psi&-\sin \psi\\0&\sin \psi&\ \ \ \ \cos \psi\end{pmatrix}\begin{pmatrix}-a \cos \alpha\\ \color{red}{+}a \sin \alpha\\0 \end{pmatrix}-\begin{pmatrix}b/2\\0\\0\end{pmatrix} $$

$$p_4=\begin{pmatrix}1&0&0\\0&\ \ \ \cos \psi& \sin \psi\\0&-\sin \psi&\cos \psi\end{pmatrix}\begin{pmatrix}-c \cos \beta\\ \color{red}{+}c \sin \beta\\0 \end{pmatrix}+\begin{pmatrix}b/2\\0\\0\end{pmatrix} $$

It remains to expand the previous expressions and then use the distance formula:

$$\delta=\sqrt{(x_1-x_4)^2+(y_1-y_4)^2+(z_1-z_4)^2}$$

$\endgroup$
4
  • $\begingroup$ Any comment ?... $\endgroup$
    – Jean Marie
    Commented Jun 18, 2021 at 18:09
  • $\begingroup$ Thank you for your answer! I think it makes sense. However, I implemented it in Java, to use for my project, and it seems it is not giving me the right values in practice. One example from practice: $a = 1.229, b = 1.522, c = 1.449, \alpha = 120.4, \beta = 110.1$ and finally $\phi = 0$. Solving the matrix operations and computing the distance gives me $\sigma = 1.43$ (I also did the matrix operations by hand, and I also got $1.43$). However, if I solve the instance by hand in another way (since $\phi = 0$, all I have to do is solve a quadrilateral), I find that $\sigma$ should be $2.665$. $\endgroup$
    – Simon H
    Commented Jun 21, 2021 at 12:32
  • $\begingroup$ I think the error might be in the matrices describing the positions of $p1$ and $p4$ when $\phi = 0$. Shouldn't we use $+ a \cos \alpha$, $+ a \sin \alpha$, $ c \cos \beta$ and $-c \sin \beta$? If $\alpha > 90$, the x-coordinate of $p1$ should be lower, if $\alpha < 90$, the x-coordinate of $p1$ should be larger. $\endgroup$
    – Simon H
    Commented Jun 22, 2021 at 12:04
  • $\begingroup$ @Simon H see the two sign rectifications I have brought (in red). $\endgroup$
    – Jean Marie
    Commented Jun 23, 2021 at 18:53
1
$\begingroup$

Take projection of A on foldline as origin. If supplements of $ \alpha, \beta \text{ are }\alpha', \beta'\text{ then } $ after projecting in three mutually perpendicular Cartesian planes directly by 3d Pythagoras

enter image description here

$$ A:( a \sin \alpha' \cos \phi, 0,a \sin \alpha' \sin \phi) $$ $$B:(c \ sin \beta', a \cos \alpha'+b + c \cos \beta',0) $$ $$L=\sqrt{ (x_B-x_A)^2+ (y_B-y_A)^2+ (z_B-z_A)^2}.$$

$\endgroup$
5
  • $\begingroup$ Thanks for your answer. What if $\alpha$ or $\beta < 90$ degrees? I implemented this answer in Java aswell, and comparing to the implementation of the answer of Jean Marie the results are the same except for when $\alpha$ or $\beta < 90$ degrees. $\endgroup$
    – Simon H
    Commented Jun 24, 2021 at 8:35
  • $\begingroup$ The formula is general for directions you have shown. Did you take supplementary angles and proper trig signs? $\endgroup$
    – Narasimham
    Commented Jun 24, 2021 at 12:06
  • $\begingroup$ I am not sure what you mean by proper trig signs. I did take supplementary angles. Actually, I do not think the problem is the $\alpha$ or $\beta$. The formula seems to work for $\phi = 0$ and $\phi = 180$. However, tried some values in between, and sometimes for two angles $\phi_1, \phi_2, \phi_1 < \phi_2$, it seems that the resulting distance $L_1 > L_2$. However, if $\phi_1 < \phi_2$ I think that $L_2$ should always be larger than $L_1$. $\endgroup$
    – Simon H
    Commented Jun 24, 2021 at 12:26
  • $\begingroup$ Apologies, a typo I noticed and corrected just now. Ok now? $\endgroup$
    – Narasimham
    Commented Jun 24, 2021 at 13:24
  • $\begingroup$ Yes, now it works perfectly. I was already a bit confused about the z-coordinates always being 0. Now it gives exactly the same solutions as the other answer! Thank you! $\endgroup$
    – Simon H
    Commented Jun 24, 2021 at 13:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .