Visual walkthrough — Reference frames — body frame, inertial frame; rotation between them

Step 1 — What is a component, really?

Look at the picture. The amber arrow $\vec{v}$ never moves. We drop two dashed lines straight down onto the two axes. Where they land gives two numbers:

$\vec{v}=\underset{\text{how far along }\hat{x}}{\underbrace{v_x}}\hat{x}+\underset{\text{how far along }\hat{y}}{\underbrace{v_y}}\hat{y}$

• $\hat{x},\hat{y}$ are unit vectors — arrows of length exactly $1$ pointing along each axis. The little hat $\hat{}$ means "length one".
• $v_x$ is the shadow of $\vec{v}$ on the $x$-axis; $v_y$ is its shadow on the $y$-axis.

Step 2 — The dot product: a machine that measures "how much along"

Step 3 — Two frames on the same arrow

Notice in the figure that the arrow is drawn once, but it has two shadow-pairs: white shadows on the white axes and cyan shadows on the cyan axes.

$\vec{v}=\underset{\text{stars’ description}}{\underbrace{v_X^{\mathcal{I}}\hat{X}+v_Y^{\mathcal{I}}\hat{Y}}}=\underset{\text{vehicle’s description}}{\underbrace{v_x^{\mathcal{B}}{\hat{x}}_{\mathcal{B}}+v_y^{\mathcal{B}}{\hat{y}}_{\mathcal{B}}}}$

• The superscript $\mathcal{I}$ tags numbers "as seen in the inertial frame".
• The superscript $\mathcal{B}$ tags numbers "as seen in the body frame".
• The middle equals sign is the whole point: two lists of numbers, one arrow.

Step 4 — Extract one body component using the dot-product trick

Now substitute the arrow written in inertial pieces, $\vec{v}=v_X^{\mathcal{I}}\hat{X}+v_Y^{\mathcal{I}}\hat{Y}+v_Z^{\mathcal{I}}\hat{Z}$, and use that the dot product distributes over the sum:

$v_x^{\mathcal{B}}=\underset{\text{cosine of angle between }{\hat{x}}_{\mathcal{B}}\text{ and }\hat{X}}{\underbrace{({\hat{x}}_{\mathcal{B}}\cdot \hat{X})}}{v}_X^{\mathcal{I}}+\underset{\text{cosine }{\hat{x}}_{\mathcal{B}},\hat{Y}}{\underbrace{({\hat{x}}_{\mathcal{B}}\cdot \hat{Y})}}{v}_Y^{\mathcal{I}}+\underset{\text{cosine }{\hat{x}}_{\mathcal{B}},\hat{Z}}{\underbrace{({\hat{x}}_{\mathcal{B}}\cdot \hat{Z})}}{v}_Z^{\mathcal{I}}$

• Each bracket is a dot product of two unit vectors → by Step 2 it is a pure cosine of the angle between an inertial axis and the body-$x$ axis.
• The picture shows exactly this: the cyan ${\hat{x}}_{\mathcal{B}}$ arrow makes an angle with each white axis, and the cosine of each angle is a weight.
• So $v_x^{\mathcal{B}}$ is a weighted blend of the three inertial numbers. The weights are cosines. These weights are the first row of the matrix.

Step 5 — Stack all three rows: the Direction Cosine Matrix

$\left[\begin{array}{c}{v}_x^{\mathcal{B}}\\ v_y^{\mathcal{B}}\\ v_z^{\mathcal{B}}\end{array}\right]=\underset{R_{\mathcal{B}\mathcal{I}}\text{— every entry is a cosine}}{\underbrace{\left[\begin{array}{ccc}{\hat{x}}_{\mathcal{B}}\cdot \hat{X}& {\hat{x}}_{\mathcal{B}}\cdot \hat{Y}& {\hat{x}}_{\mathcal{B}}\cdot \hat{Z}\\ {\hat{y}}_{\mathcal{B}}\cdot \hat{X}& {\hat{y}}_{\mathcal{B}}\cdot \hat{Y}& {\hat{y}}_{\mathcal{B}}\cdot \hat{Z}\\ {\hat{z}}_{\mathcal{B}}\cdot \hat{X}& {\hat{z}}_{\mathcal{B}}\cdot \hat{Y}& {\hat{z}}_{\mathcal{B}}\cdot \hat{Z}\end{array}\right]}}\left[\begin{array}{c}{v}_X^{\mathcal{I}}\\ v_Y^{\mathcal{I}}\\ v_Z^{\mathcal{I}}\end{array}\right]$

• Row 1 was born from dotting with ${\hat{x}}_{\mathcal{B}}$ — so row $i$ is body axis $i$ written in inertial coordinates.
• Column $j$ carries the inertial-$j$ number into all three body slots.
• Every single entry is $\cos (\text{angle between two axes})$ → the name Direction Cosine Matrix.

$\boxed{{\vec{v}}^{\mathcal{B}}=R_{\mathcal{B}\mathcal{I}}{\vec{v}}^{\mathcal{I}}}$

• Read the subscripts right-to-left: start in $\mathcal{I}$, land in $\mathcal{B}$. The $\mathcal{I}$ "cancels" against the $\mathcal{I}$ of the vector — like units cancelling.

Step 6 — Make it concrete: a pure spin about $z$

In the top-down view, the cyan body axes are rotated $+\theta$ from the white inertial axes. Measuring the four in-plane angles:

• ${\hat{x}}_{\mathcal{B}}\cdot \hat{X}=\cos \theta$ (angle $\theta$ between them)
• ${\hat{x}}_{\mathcal{B}}\cdot \hat{Y}=\cos (90^{\circ }-\theta )=\sin \theta$
• ${\hat{y}}_{\mathcal{B}}\cdot \hat{X}=\cos (90^{\circ }+\theta )=-\sin \theta$ ← the crucial minus sign
• ${\hat{y}}_{\mathcal{B}}\cdot \hat{Y}=\cos \theta$

$R_z(\theta )=\left[\begin{array}{ccc}\cos \theta & \sin \theta & 0\\ -\sin \theta & \cos \theta & 0\\ 0& 0& 1\end{array}\right]$

• The bottom row and right column are $0,0,1$ because $\hat{z}$ is shared — it makes a $90^{\circ }$ angle with both in-plane axes ($\cos 90^{\circ }=0$) and $0^{\circ }$ with itself ($\cos 0^{\circ }=1$).
• The lone minus on $-\sin \theta$ is exactly the picture: the new ${\hat{y}}_{\mathcal{B}}$ leans into $-\hat{X}$, so its cosine with $\hat{X}$ is negative. That minus is geometry, not a convention you must memorise.

Step 7 — The degenerate case $\theta =0$: the identity

$R_z(0)=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right]=I$

• Body axes sit exactly on inertial axes: each axis is perfectly aligned with its twin ($\cos 0^{\circ }=1$ on the diagonal) and perpendicular to the others ($\cos 90^{\circ }=0$ off-diagonal).
• Applying $I$ changes no number: ${\vec{v}}^{\mathcal{B}}={\vec{v}}^{\mathcal{I}}$. Sanity anchor — every correct rotation matrix must collapse to $I$ at zero angle.

Step 8 — Going the other way is just a flip (the transpose)

Entry $(i,j)$ of $R{R}^{\mathsf{T}}$ is (body axis $i$) $\cdot$ (body axis $j$). Since body axes are mutually perpendicular unit vectors (orthonormal):

${\hat{e}}_i^{\mathcal{B}}\cdot {\hat{e}}_j^{\mathcal{B}}=\{\begin{array}{ll}1& i=j(\text{same axis, }\cos 0^{\circ })\\ 0& i\ne j(\text{perpendicular, }\cos 90^{\circ })\end{array}$

$\Rightarrow R{R}^{\mathsf{T}}=I\Rightarrow \boxed{R^{-1}=R^{\mathsf{T}}}\Rightarrow {\vec{v}}^{\mathcal{I}}=R_{\mathcal{B}\mathcal{I}}^{\mathsf{T}}{\vec{v}}^{\mathcal{B}}$

• Why it matters onboard: the flight computer never runs an expensive matrix inversion. Reversing a frame conversion costs one transpose — literally swapping rows and columns. This is the IMU-to-star pipeline's cheapest, most reliable step.
• This is also the resolution of the classic slip warned about in Euler angles and gimbal lock and Quaternions for attitude representation: "rotate the vector" versus "rotate the frame" differ by exactly this transpose.

The one-picture summary

One arrow, two frames, one table of cosines connecting them — and the reverse trip is the same table flipped across its diagonal.

${\vec{v}}^{\mathcal{B}}=R_{\mathcal{B}\mathcal{I}}{\vec{v}}^{\mathcal{I}}\text{and}{\vec{v}}^{\mathcal{I}}=R_{\mathcal{B}\mathcal{I}}^{\mathsf{T}}{\vec{v}}^{\mathcal{B}}$

Reveal checks: