Modified Rodrigues parameters — singularity-free, compact

WHY do we even need MRPs?

We already have several ways to store orientation. Each has a flaw:

• Direction Cosine Matrix (DCM): 9 numbers, 6 constraints → wasteful, awkward to integrate.
• Euler angles: 3 numbers, but gimbal lock — a singularity at $\pm 90^\circ$ where two axes align and you lose a degree of freedom.
• Quaternions: 4 numbers, no singularity ever, but they carry a redundant constraint $\|q\|=1$ and a sign ambiguity ($q$ and $-q$ are the same rotation).

WHAT are MRPs? — Derivation from the axis–angle

Every rotation = rotate by angle $\Phi$ about unit axis $\hat{e}$. Start from the Euler symmetric parameters (quaternion):

$q_0 = \cos\frac{\Phi}{2}, \qquad \mathbf{q} = \hat{e}\,\sin\frac{\Phi}{2}$

The classical Rodrigues parameters (Gibbs vector) are $\mathbf{g} = \hat{e}\tan(\Phi/2) = \mathbf{q}/q_0$. These blow up at $\Phi = 180^\circ$ (where $q_0=0$). Too early.

Trick: use the half-angle again. Define MRPs by dividing by $(1+q_0)$ instead of $q_0$:

$\boxed{\;\boldsymbol{\sigma} = \dfrac{\mathbf{q}}{1+q_0}\;}$

Why does $\mathbf{q}/(1+q_0)$ equal $\hat{e}\tan(\Phi/4)$? Substitute the quaternion: $\frac{\mathbf{q}}{1+q_0} = \frac{\hat{e}\sin(\Phi/2)}{1+\cos(\Phi/2)} = \hat{e}\,\tan\!\frac{\Phi}{4}$ Why this step? We used the identity $\dfrac{\sin\theta}{1+\cos\theta} = \tan\dfrac{\theta}{2}$ with $\theta = \Phi/2$. The half-angle of the half-angle = quarter-angle. That extra halving is exactly what pushes the singularity from $180^\circ$ (Gibbs) out to $360^\circ$.

The shadow set — how we kill the last singularity

Because $q$ and $-q$ are the same physical rotation, there are two MRP vectors for every attitude:

$\boldsymbol{\sigma} = \frac{\mathbf{q}}{1+q_0}, \qquad \boldsymbol{\sigma}^S = \frac{-\mathbf{q}}{1-q_0} = -\frac{\boldsymbol{\sigma}}{\|\boldsymbol{\sigma}\|^2}$

Why this works: $\|\boldsymbol{\sigma}\| = \tan(\Phi/4)$. At $\Phi = 180^\circ$, $\|\boldsymbol{\sigma}\|=1$; at $\Phi\to 360^\circ$, $\|\boldsymbol{\sigma}\|\to\infty$. The shadow set maps $\Phi \mapsto 360^\circ - \Phi$, so it always has $\|\boldsymbol{\sigma}^S\|<1$ exactly when the original exceeds 1. Keeping $\|\boldsymbol{\sigma}\|\le 1$ guarantees you never approach the singularity.

The kinematic differential equation (HOW to propagate)

Given body angular velocity $\boldsymbol{\omega}$, the MRPs evolve as:

$\dot{\boldsymbol{\sigma}} = \frac{1}{4}\,B(\boldsymbol{\sigma})\,\boldsymbol{\omega}$

$B(\boldsymbol{\sigma}) = \big(1-\sigma^2\big)\,\mathbb{I}_3 + 2[\boldsymbol{\sigma}^\times] + 2\,\boldsymbol{\sigma}\boldsymbol{\sigma}^{\!\top}, \qquad \sigma^2 = \boldsymbol{\sigma}^{\!\top}\boldsymbol{\sigma}$

where $[\boldsymbol{\sigma}^\times]$ is the cross-product (skew-symmetric) matrix.

A useful property (Steel-man): $B(\boldsymbol{\sigma})$ is nearly orthogonal: $B^{\!\top}B = \big(1+\sigma^2\big)^2\,\mathbb{I}_3 \;\Rightarrow\; B^{-1} = \frac{B^{\!\top}}{(1+\sigma^2)^2}.$ This makes inverting the kinematics (solving for $\boldsymbol{\omega}$) trivial — a big practical win for control law design.

Worked examples

Common mistakes

Flashcards

Connections

• Quaternions (Euler symmetric parameters) — parent representation; $\boldsymbol{\sigma}=\mathbf q/(1+q_0)$
• Classical Rodrigues parameters (Gibbs vector) — sibling, $\tan(\Phi/2)$, singular at $180^\circ$
• Euler angles and Gimbal Lock — the singularity problem MRPs improve on
• Direction Cosine Matrix (DCM) — the 9-number global standard
• Attitude kinematics and $\boldsymbol\omega$ — where $\dot{\boldsymbol\sigma}=\frac14B\boldsymbol\omega$ lives
• Spacecraft attitude control laws — MRP feedback controllers exploit $B^{-1}$ simplicity

Concept Map

Hinglish (regional understanding)