Visual walkthrough — Gimbal lock — problem with Euler angles at θ = ±90°

3.5.5 · D2 · Physics › Guidance, Navigation & Control (GNC) › Gimbal lock — problem with Euler angles at θ = ±90°

---

Step 1 — Teen axes, teen spins (vocabulary)

KYA HAI. Ek orientation ka matlab hai "object kis taraf jhuka hua hai." Hum isse teen simple turns se banate hain, har ek ek right-handed frame ke ek axis ke baare mein. Right-handed frame matlab teen arrows jo ek doosre se $90°$ par hain, labelled $x$ (forward), $y$ (left), $z$ (up), is tarah arrange kiye hue ki right hand ko $x$ se $y$ ki taraf curl karne par thumb $z$ ki taraf point kare.

YEH TEEN KYUN. 3D mein koi bhi tilt exactly teen independent turns stack karke reach ki ja sakti hai — na zyada, na kam. Woh number teen "degrees of freedom" (DOF) hai: teen independent dials. Ek ko khona hi is page ki poori kahani hai.

PICTURE. Figure s01 mein teen named turns dikhaye gaye hain:

• $\psi$ (yaw) — up-axis $z$ ke baare mein spin. Socho "left/right turn."
• $\theta$ (pitch) — left-axis $y$ ke baare mein spin. Socho "nose up/down."
• $\varphi$ (roll) — forward-axis $x$ ke baare mein spin. Socho "barrel roll."

---

Step 2 — Ek spin ko grid mein badalna (rotation matrix)

KYA HAI. In turns ke saath compute karne ke liye hum har ek ko numbers ki $3\times 3$ box se replace karte hain jise rotation matrix kehte hain. Yahan matrix bas ek bookkeeping grid hai: isko arrow "before" do, woh arrow "after" deta hai.

SIRF ANGLE NAHI, MATRIX KYUN. Kyunki humein turns stack karne hote hain, aur stacked turns ek fixed order mein grids ki tarah multiply hote hain. Angles akele chain nahi ho sakte; grids ho sakte hain. Dekho Rotation matrices SO(3) jis group mein yeh rehte hain.

PICTURE. Figure s02 pitch turn $R_y(\theta )$ leta hai aur dikhata hai ki har entry ek test arrow ke saath kya karta hai.

$$R_y(\theta )=\left(\begin{array}{ccc}\underset{\text{keep-forward}}{\underbrace{\cos \theta }}& 0& \underset{\text{tip up}}{\underbrace{\sin \theta }}\\ 0& \underset{\text{left axis frozen}}{\underbrace{1}}& 0\\ \underset{\text{tip forward-from-up}}{\underbrace{-\sin \theta }}& 0& \underset{\text{keep-up}}{\underbrace{\cos \theta }}\end{array}\right)$$

Har symbol, jahan woh baitha hai wahan:

• Diagonal par $\cos \theta$ $=$ "purani direction kitni bachti hai."
• Off-diagonal $\sin \theta$ $=$ "forward se up mein (ya back mein) kitna leak hota hai."
• Middle row mein akela $1$ $=$ pitch axis $y$ untouched hai — woh khud hi spin axis hai.

Doosre do dials, usi tarah likhay. Taki koi bhi grid guess karne ko na baache, yahan yaw aur roll matrices puri tarah likhi hain. Har ek apna spin axis frozen rakhti hai (us axis par akela $1$) aur baaki do components ko $\cos$ (keep) aur $\sin$ (leak) ke through mix karti hai:

$$R_z(\psi )=\left(\begin{array}{ccc}\underset{\text{keep}}{\underbrace{\cos \psi }}& -\sin \psi & 0\\ \underset{\text{leak}}{\underbrace{\sin \psi }}& \cos \psi & 0\\ 0& 0& \underset{z\text{frozen}}{\underbrace{1}}\end{array}\right),R_x(\varphi )=\left(\begin{array}{ccc}\underset{x\text{frozen}}{\underbrace{1}}& 0& 0\\ 0& \underset{\text{keep}}{\underbrace{\cos \varphi }}& -\sin \varphi \\ 0& \underset{\text{leak}}{\underbrace{\sin \varphi }}& \cos \varphi \end{array}\right)$$

• $R_z(\psi )$ $z$ (yaw axis) ko freeze karta hai aur $x,y$ ko mix karta hai.
• $R_x(\varphi )$ $x$ (roll axis) ko freeze karta hai aur $y,z$ ko mix karta hai.

---

Step 3 — Middle dial ko danger value par freeze karo

KYA HAI. Hum pitch ko uske suspected singular value $\theta =+90°$ par set karte hain aur dekhte hain ki pitch grid kya ban jaata hai. $90°$ par: $\cos 90°=0$ aur $\sin 90°=1$.

YAHAN KYUN. Middle turn hi ek aisa hai jo last axis ko first ke relative tilt karta hai. Isse $90°$ tak push karo aur hum suspect karte hain ki last axis ($x$, roll) exactly first axis ($z$, yaw) par lie karega. Hum is suspicion ko directly test karte hain na ki guess karte hain.

PICTURE. Figure s03: forward-arrow $x$ ko swing hote hue dekho jab tak woh $z$ ke along point na kare. Dono "leak" terms $\sin \theta$ apne maximum $1$ tak pahunch jaate hain; dono "keep" terms $\cos \theta$ $0$ par collapse ho jaate hain.

$$R_y(90°)=\left(\begin{array}{ccc}\underset{\text{no forward kept}}{\underbrace{0}}& 0& \underset{\text{fully tipped up}}{\underbrace{1}}\\ 0& 1& 0\\ \underset{\text{up}\to \text{back}}{\underbrace{-1}}& 0& \underset{\text{no up kept}}{\underbrace{0}}\end{array}\right)$$

Key visual: purani forward direction ab seedha upar khadi hai. Yahi woh alignment hai jo yaw aur roll ko merge kar degi.

---

Step 4 — Multiply karo aur collapse padho

KYA HAI. Hum Steps 2–3 ki teen grids se $R=R_z(\psi )R_y(90°)R_x(\varphi )$ compute karte hain, aur result ko ghoorte hain:

$$R=\left(\begin{array}{ccc}0& \sin (\varphi -\psi )& \cos (\varphi -\psi )\\ 0& \cos (\varphi -\psi )& -\sin (\varphi -\psi )\\ -1& 0& 0\end{array}\right)$$

YEH PUNCHLINE KYUN HAI. Har non-trivial entry dekho: unmein $\psi$ aur $\varphi$ sirf single combination $(\varphi -\psi )$ ke through hain. Toh agar tum yaw dial ko $\delta$ se upar karo aur roll dial ko bhi $\delta$ se, difference $\varphi -\psi$ unchanged rahta hai — aur isliye $R$ unchanged rahta hai. Do dials, ek net effect: ek degree of freedom gayab ho gayi.

PICTURE. Figure s04 entries ko plot karta hai jab $\psi$ aur $\varphi$ dono vary karte hain. Surface diagonal $\varphi =\psi +\text{const}$ ke along flat hai — $(\psi ,\varphi )$ pairs ki poori valley same orientation deti hai.

---

Step 5 — Doosra pole, aur beech ka sab kuch

KYA HAI. Hume sab cases cover karne hain, sirf $+90°$ nahi. Teen regimes:

• $\theta =-90°$: same collapse, opposite sign. $\cos \theta =0$, $\sin \theta =-1$ set karo, toh pitch grid ban jaata hai $R_y(-90°)=\left(\begin{array}{ccc}0& 0& -1\\ 0& 1& 0\\ 1& 0& 0\end{array}\right)$. $R_z(\psi )R_y(-90°)R_x(\varphi )$ multiply karne par neeche diya collapse milta hai.
• $-90°<\theta <+90°$ (safe interior): $\cos \theta \ne 0$, dono axes genuinely ek doosre se tilt hain, teeno dials independently kaam karte hain. Full 3 DOF.
• $\theta$ near $\pm 90°$ par exactly nahi: safe nahi, jaise Step 6 dikhata hai.

SIGN KYUN FLIP HOTA HAI. $+90°$ par forward axis seedha upar $+z$ ki taraf point kar raha tha, toh roll aur yaw ek hi direction mein chale aur unke effects subtract hue ($\varphi -\psi$). $-90°$ par forward axis seedha neeche $-z$ ki taraf point karta hai — roll ab yaw ke opposite direction mein chalta hai, toh unke effects add hote hain ($\varphi +\psi$). Akela sign change $\sin \theta :+1\to -1$ hi woh hai jo multiplication ke through $-\psi \to +\psi$ le jaata hai. Explicitly:

$$R{|}_{\theta =-90°}=\left(\begin{array}{ccc}0& -\sin (\varphi +\psi )& -\cos (\varphi +\psi )\\ 0& \cos (\varphi +\psi )& -\sin (\varphi +\psi )\\ 1& 0& 0\end{array}\right)$$

— har non-trivial entry ab sirf $(\varphi +\psi )$ par depend karti hai.

PICTURE. Figure s05 teen regimes ko side by side line up karta hai: yaw axis aur roll axis fanned apart (interior), merged pointing up ($+90°$), merged pointing down ($-90°$).

---

Step 6 — Rates kyun explode karte hain (ek point nahi, neighborhood)

KYA HAI. Orientation static nahi hoti; ek GNC computer body spin rates $(p,q,r)$ measure karta hai — jahan $p$ roll axis $x$ ke baare mein spin rate hai, $q$ pitch axis $y$ ke baare mein, aur $r$ yaw axis $z$ ke baare mein (Step 1 ke wahi teen axes) — aur unhe Euler angles mein integrate karta hai. Bridge (Angular velocity kinematics se) yeh hai:

$$\left(\begin{array}{c}\dot{\varphi }\\ \dot{\theta }\\ \dot{\psi }\end{array}\right)=\left(\begin{array}{ccc}1& \sin \varphi \tan \theta & \cos \varphi \tan \theta \\ 0& \cos \varphi & -\sin \varphi \\ 0& \frac{\sin \varphi }{\cos \theta }& \frac{\cos \varphi }{\cos \theta }\end{array}\right)\left(\begin{array}{c}p\\ q\\ r\end{array}\right)$$

Do symbols culprits hain, yahan define kiye gaye:

• $\tan \theta =\frac{\sin \theta }{\cos \theta }$ — "steepness"; yeh jawab deta hai ki tilt kitni tezi se badhti hai per unit turn ke. Jab $\theta \to 90°$, denominator $\cos \theta \to 0$, toh $\tan \theta \to \infty$.
• $\frac{1}{\cos \theta }$ — same vanishing denominator, same blow-up.

YEH EK REGION KYUN HAI, POINT NAHI. Kyunki $1/\cos \theta$ $90°$ ke aas-paas angles ki ek band ke liye bada hota hai, sirf exact value par nahi. $\theta =89.9°$ par, $\cos \theta \approx 0.001745$, toh $1/\cos \theta \approx 573$. $r$ mein sensor noise ki ek whisper $\dot{\psi }$ mein ek scream ban jaati hai.

PICTURE. Figure s06 $1/\cos \theta$ ko $\theta$ ke across plot karta hai: zyaatar range ke liye ek smooth curve jo $\theta \to \pm 90°$ par vertical asymptote ki taraf rocket karta hai. Unstable band shaded hai.

---

Step 7 — Physics kabhi nahi tuti; sirf chart tuta

KYA HAI. Rigid body ab bhi har direction mein perfectly freely rotate karta hai. Jo fail hua woh hai coordinate chart — woh particular tarika jis mein $(\varphi ,\theta ,\psi )$ orientations ko label karta hai. Yahi woh hai jo Singularities of coordinate charts describe karta hai: map interior mein theek hai, poles par singular hai, jaise globe ke North Pole par meridian lines sab crash ho jaati hain.

FIX KYUN KAAM KARTA HAI. Quaternions poore $SO(3)$ ko smoothly wrap karte hain bina kisi singular orientation ke, isliye woh kabhi $\cos \theta$ se divide nahi karte. Unka akela quirk ($q\equiv -q$) ek lost DOF nahi hai.

PICTURE. Figure s07: pole par crowding meridians wala ek globe (Euler singularity) aur ek smooth quaternion cover jo bina kisi pinch point ke hai, side by side.

---

Ek-picture summary

KYA HAI. Figure s08 poore walkthrough ko compress karta hai: interior pitch yaw aur roll ko fanned apart rakhta hai (3 DOF); jab pitch $\pm 90°$ ki taraf swing karta hai toh roll axis yaw axis par chadh jaata hai, matrix sirf $(\varphi \mp \psi )$ par depend karta hai, aur rate curve $1/\cos \theta$ chhath phaadk ke oopar chali jaati hai. Ek nazar poori kahani bata deti hai.

Recall Feynman retelling — poora walkthrough simple shabdon mein

Ek toy plane socho jismein teen dials hain: turn (yaw), nose up/down (pitch), barrel-roll (roll). Humne har dial ko numbers ki ek chhoti grid ki tarah likha taki hum unhe stack kar sakein (Steps 1–2). Phir humne middle dial — pitch — ko crank kiya jab tak nose seedha ceiling ki taraf point na kar de (Step 3). Jab humne teeno grids multiply ki, kuch spooky hua: answer sirf turn dial aur roll dial ke difference ki parwah karta tha (Step 4). Dono dials ko same amount se turn karo aur plane bilkul nahi hilda — do dials ek mein fuse ho gaye the. Wahi trap nose-straight-down par bhi hai, sirf wahan dials subtract ki jagah add karte hain (Step 5). Aur bura yeh hai ki woh equation jo real spin ko dial-motion mein convert karti hai usme $1/\cos \theta$ hai, jo poles ke paas almost-zero se division ban jaata hai, toh sensor mein ek tiny wobble computer ko insane spin command karne par majboor kar deta hai (Step 6). Lekin plane khud kabhi stuck nahi tha — sirf hamare teen dials confuse ho gaye. Dials ko quaternions se replace karo, jinka map kabhi pinch nahi hota, aur confusion gayab ho jaata hai (Step 7).

---

Connections

• Euler angles — woh three-dial representation jise humne dissect kiya
• Rotation matrices SO(3) — woh grids jo humne Steps 2–4 mein multiply ki
• Angular velocity kinematics — $1/\cos \theta$ rate blow-up ka source (Step 6)
• Quaternions — pinch-free fix (Step 7)
• Singularities of coordinate charts — woh general reason ki chart kyun fail hota hai jabki physics theek hai
• Attitude determination and control — jahan yeh real GNC mein daata hai
• Apollo Guidance Computer — historic "gimbal lock" alarm