1. Why Eigenaxis Rotation Is Special

A rigid-body attitude change is not unique: you can reach the same final orientation using many different sequences and mixtures of rotations. A spacecraft could:

• rotate about one axis and then another,
• execute a composite multi-axis maneuver,
• follow a curved path in attitude space, or
• rotate about a single fixed axis from start to finish.

Among all those possibilities, there is exactly one rotation that minimizes the total angular displacement. That minimal displacement motion is:

\[ \text{A single rotation about one fixed axis by the smallest possible angle.} \]

The distinguished axis is the eigenaxis of the relative attitude. This result is purely geometric: it does not require knowledge of actuator limits, torque saturation, control gains, or feedback law. It arises from the structure of the rotation group $SO(3)$ itself.

What D.3 builds

• Minimum-angle rotation geometry: the unique minimal rotation between two attitudes.
• Axis–angle representation: how $(\hat{\mathbf e},\theta)$ parameterizes the relative rotation.
• Quaternion shortest-path structure: how $q_4 \ge 0$ selects the smaller rotation.
• Optimality conditions: when “shortest angle” also becomes “shortest time.”
• Comparison: why multi-axis maneuvers may be faster for asymmetric inertia even if they rotate more.

2. Minimum-Angle Rotation Between Two Attitudes

Let the spacecraft’s attitude be represented by a direction cosine matrix (DCM):

\[ \mathbf R_c \in SO(3) \quad \text{(current attitude)}, \qquad \mathbf R_d \in SO(3) \quad \text{(desired attitude)}. \]

The rotation that maps the current attitude to the desired attitude is the relative rotation:

\[ \mathbf R_e = \mathbf R_d \mathbf R_c^{T}. \]

By construction, this satisfies:

\[ \mathbf R_d = \mathbf R_e \mathbf R_c. \]

Since $\mathbf R_e \in SO(3)$, it must represent a proper rotation. A fundamental result from rigid-body kinematics states:

Therefore $\mathbf R_e$ can be written in axis–angle (Rodrigues) form:

\[ \mathbf R_e = \mathbf I + \sin\theta \,[\hat{\mathbf e}]_\times + (1-\cos\theta)\,[\hat{\mathbf e}]_\times^2, \]

• $\hat{\mathbf e} \in \mathbb{R}^3$ is a unit rotation axis (the eigenaxis).
• $\theta \in [0,\pi]$ is the rotation angle (chosen minimal).
• $[\hat{\mathbf e}]_\times$ is the skew-symmetric cross-product matrix.

The skew-symmetric matrix is:

\[ [\hat{\mathbf e}]_\times = \begin{bmatrix} 0 & -e_3 & e_2\\ e_3 & 0 & -e_1\\ -e_2 & e_1 & 0 \end{bmatrix}. \]

Why it is called the eigenaxis

The eigenaxis is the direction left unchanged by the rotation:

\[ \mathbf R_e \hat{\mathbf e} = \hat{\mathbf e}. \]

This means $\hat{\mathbf e}$ is an eigenvector of $\mathbf R_e$ with eigenvalue 1. Geometrically:

• the axis direction is invariant under the rotation,
• all directions orthogonal to it rotate through the angle $\theta$.

3. Axis–Angle Representation

The axis–angle form gives a compact representation of the entire attitude change. The angle is obtained from the trace identity:

\[ \theta = \cos^{-1}\left(\frac{\mathrm{trace}(\mathbf R_e)-1}{2}\right). \]

Once $\theta$ is known, the eigenaxis can be extracted from the antisymmetric part of $\mathbf R_e$:

\[ \hat{\mathbf e} = \frac{1}{2\sin\theta} \begin{bmatrix} R_{32}-R_{23}\\ R_{13}-R_{31}\\ R_{21}-R_{12} \end{bmatrix}. \]

Therefore the attitude change is described as:

\[ \boxed{\text{Rotate about } \hat{\mathbf e} \text{ by angle } \theta.} \]

Why this is the minimum-angle rotation

• For a fixed pair $(\mathbf R_c,\mathbf R_d)$, the relative rotation $\mathbf R_e$ is fixed.
• Choosing $\theta \in [0,\pi]$ selects the smaller of the two equivalent rotations.
• This produces the unique rotation that minimizes angular displacement (the geodesic on $SO(3)$).

4. Quaternion-Based Shortest Rotation

Axis–angle geometry is beautifully encoded by quaternions. Let the attitude error quaternion be:

\[ \mathbf q_e = (q_1, q_2, q_3, q_4), \]

using the scalar-last convention. Axis–angle parameters relate to quaternion components as:

\[ q_4 = \cos\left(\frac{\theta}{2}\right), \qquad \mathbf q_v = \hat{\mathbf e}\,\sin\left(\frac{\theta}{2}\right), \]

where $\mathbf q_v = (q_1,q_2,q_3)$ is the vector part. Hence:

\[ \theta = 2\cos^{-1}(q_4), \qquad \hat{\mathbf e} = \frac{\mathbf q_v}{\sin(\theta/2)}. \]

Shortest-path logic in quaternions

Quaternions double-cover $SO(3)$:

\[ \mathbf q \equiv -\mathbf q. \]

Both represent the same physical attitude, but they imply different angles:

• If $q_4 > 0$, then $\theta \in [0,\pi]$ (the short rotation).
• If $q_4 < 0$, the implied rotation exceeds $\pi$ (the long way around).

Therefore the geometric minimum rotation is enforced by the sign choice:

\[ \boxed{q_4 \ge 0.} \]

This selects the quaternion on the “short” hemisphere of the unit quaternion sphere $S^3$ and guarantees the minimum-angle representation.

5. Eigenaxis Rotation Dynamics (Pure Geometric Case)

Consider a motion in which the spacecraft rotates only about the eigenaxis direction:

\[ \boldsymbol{\omega}(t) = \omega_e(t)\,\hat{\mathbf e}. \]

In this case:

• the direction of angular velocity remains fixed (collinear with $\hat{\mathbf e}$),
• the quaternion vector part remains aligned with $\hat{\mathbf e}$,
• no transverse components are generated by the kinematics themselves.

The rotation collapses to a scalar evolution in the rotation angle:

\[ \dot{\theta} = \omega_e. \]

This is the key “elegance” of eigenaxis motion: the full 3D attitude kinematics reduce to a single scalar differential equation when the motion stays on the eigenaxis.

6. When Is Eigenaxis Rotation Optimal?

Case 1 — Kinematic optimality (always)

Eigenaxis rotation always minimizes angular displacement. No other continuous rotation between the same two attitudes yields a smaller rotation angle.

Case 2 — Time-optimal for inertially symmetric bodies

If the spacecraft is inertially symmetric:

\[ J_1 = J_2 = J_3, \]

then Euler’s rotational dynamics decouple (no gyroscopic coupling), and the time-optimal control under bounded torque takes a bang-bang form aligned with the eigenaxis:

\[ \mathbf u = u_{\max}\,\mathrm{sgn}(\theta)\,\hat{\mathbf e}. \]

The resulting ideal structure is:

• accelerate along $\hat{\mathbf e}$,
• coast if needed,
• decelerate along the same axis to reach rest-to-rest.

In this special inertia case, eigenaxis rotation is both:

• geometrically minimal, and
• dynamically time-optimal.

Case 3 — Asymmetric inertia

If:

\[ J_1 \ne J_2 \ne J_3, \]

then gyroscopic coupling terms $\boldsymbol{\omega}\times(\mathbf J\boldsymbol{\omega})$ appear in the dynamics. As a result:

• pure eigenaxis motion may not remain invariant under dynamics,
• multi-axis torque shaping may achieve shorter maneuver time even if it rotates more,
• the “shortest-angle” path is not guaranteed to be the “shortest-time” path.

7. Eigenaxis vs Multi-Axis Reorientation

Eigenaxis motion is the unique minimal geodesic in rotation space. Multi-axis maneuvers can still be useful (especially with actuator constraints or asymmetric inertia), but they do not follow the shortest geometric path.

Geometrically, eigenaxis rotation corresponds to the shortest arc on $SO(3)$. Multi-axis motion corresponds to composite rotations that “bend” the path and increase total angular displacement.

8. Geometric Visualization

8.1 Quaternion sphere picture ($S^3$)

• Unit quaternions lie on a 4D unit sphere $S^3$.
• Physical attitudes correspond to antipodal points ($\mathbf q$ and $-\mathbf q$ are identical).
• Eigenaxis rotation traces a great-circle arc on $S^3$.
• Enforcing $q_4 \ge 0$ selects the shorter hemisphere (shortest rotation).

8.2 Rotation manifold picture ($SO(3)$)

In the geometry of $SO(3)$, the eigenaxis rotation is the minimal geodesic between two orientations. That is why it is the shortest path in rotation space and why it is the natural “optimal” rotation in a purely geometric sense.

9. Connection to Quaternion Feedback (Preview)

Consider a typical quaternion feedback structure:

\[ \mathbf u = -\mathbf K\,\mathbf q_v - \mathbf C\,\boldsymbol{\omega}. \]

This drives both:

\[ \mathbf q_v \rightarrow \mathbf 0, \qquad \boldsymbol{\omega} \rightarrow \mathbf 0. \]

If the initial angular velocity direction is aligned with the initial quaternion vector part, then the closed-loop motion often stays close to a single-axis rotation. That is why many quaternion-feedback slews “look like” eigenaxis rotations in practice, even when the controller was not explicitly designed as an eigenaxis command law.

Summary: What D.3 established

• The minimum-angle reorientation between two attitudes is a single-axis rotation (eigenaxis) by $\theta \in [0,\pi]$.
• The relative attitude is $\mathbf R_e = \mathbf R_d \mathbf R_c^T$ and admits an axis–angle form via Rodrigues’ formula.
• The eigenaxis is the invariant direction satisfying $\mathbf R_e \hat{\mathbf e} = \hat{\mathbf e}$.
• The angle follows $\theta=\cos^{-1}\!\left(\frac{\mathrm{trace}(\mathbf R_e)-1}{2}\right)$ and the axis is extracted from the antisymmetric part.
• Quaternion error encodes the same geometry: $q_4=\cos(\theta/2)$ and $\mathbf q_v=\hat{\mathbf e}\sin(\theta/2)$.
• The shortest-path condition is $\boxed{q_4 \ge 0}$, selecting the “short hemisphere” on $S^3$.
• Pure eigenaxis motion reduces 3D attitude kinematics to the scalar relation $\dot{\theta}=\omega_e$.
• Eigenaxis is always geometrically minimal, and becomes time-optimal for symmetric inertia, but not necessarily for asymmetric bodies.