Foundations — Pole placement — Ackermann's formula

3.5.34 · D1 · Physics › Guidance, Navigation & Control (GNC) › Pole placement — Ackermann's formula

Yeh page assume karta hai ki aapne kuch nahi dekha. Parent note mein har ek symbol yahan se ground up banaya gaya hai, ek aisi sequence mein jahan har idea sirf usse pehle wale ideas par lean karta hai.

1. State vector $x$ — "abhi system kya hai"

Socho ek cart ek track par hai. Aage kya hoga yeh jaanne ke liye do facts chahiye: woh kahan hai aur kitni tezi se move kar raha hai. Un numbers ko ek stacked list mein bundle karo:

$x=\left[\begin{array}{c}{x}_1\\ x_2\end{array}\right]=\left[\begin{array}{c}\text{position}\\ \text{velocity}\end{array}\right].$

Picture: $x$ ek single point hai jo ek aisi space mein float kar raha hai jiske axes state variables hain. Jaise-jaise system evolve hota hai, woh point ek path trace karta hai. Is topic mein baaki sab kuch is point ko steer karne ke baare mein hai.

Topic ko kyun chahiye: pole placement design karta hai ki point kaise move karta hai, isliye pehle hume point ka naam chahiye.

2. Dot $\dot{x}$ — "change ki rate" (derivative ka pehla taste)

Yeh chhota sa dot matlab hai "yeh per second kitni tezi se change ho raha hai."

Figure mein amber arrows dekho: har ek woh jagah ka $\dot{x}$ hai. Blue curve bas arrows ko follow karti hai. Topic ko yeh isliye chahiye kyunki poora system $\dot{x}$ ke liye ek rule ke roop mein likha jaata hai.

3. Matrices $A$ aur $B$ — "woh machine jo state ko motion mein badalta hai"

Ek matrix numbers ki ek rectangular grid hai jo ek transformer ki tarah kaam karti hai: isse ek vector do, yeh doosra vector wapas deta hai. $A$ ko $x$ se multiply karna matlab hai "state variables ko fixed proportions mein combine karo taaki answer ka har component bane."

$\left[\begin{array}{cc}a& b\\ c& d\end{array}\right]\left[\begin{array}{c}{x}_1\\ x_2\end{array}\right]=\left[\begin{array}{c}a{x}_1+b{x}_2\\ c{x}_1+d{x}_2\end{array}\right].$

Grid ki row $i$ ko $x$ mein dot karke output ki entry $i$ milti hai.

Inhe milao, parent ka headline equation:

Single-column $B$ kyun: yeh page (Ackermann ki classic form ki tarah) single-input hai — exactly ek knob $u$. Dekho State feedback control.

4. Input $u$ aur feedback gain $K$ — "push aur aap ise kaise decide karte hain"

$u$ aapka ek control knob hai — ek single number jo aap har instant choose kar sakte ho (motor voltage, thrust, steering torque).

Clever move hai state feedback: $u$ haath se mat chuno — jo measure karo ussse compute karo.

$u=-Kx$ ko $\dot{x}=Ax+Bu$ mein substitute karo:

$\dot{x}=Ax-BKx=(A-BK)x.$

5. Eigenvalues $\lambda$ — "secret numbers = poles"

Kuch special vectors, jab $A$ unhe hit karta hai, same direction mein nikalta hai, sirf stretch hoke:

$Av=\lambda v.$

Yahan $v$ eigenvector hai (special direction) aur $\lambda$ uska eigenvalue hai (stretch factor). Dekho Eigenvalues and system stability.

Figure mein complex plane padho — yeh poora case list hai:

Topic kyun exist karta hai: agar poles is plane ke ek bure column mein hain, toh hum unhe stable, well-damped region mein left drag karna chahte hain. Yahi pole placement hai.

6. $e^{\lambda t}$ — "decay, growth, aur wobble ki shape"

Exponential $e^{\lambda t}$ woh ek function hai jo apni khud ki rate of change hai ($\frac{d}{dt}{e}^{\lambda t}=\lambda e^{\lambda t}$) — bilkul isliye yeh $\dot{y}=\lambda y$ solve karta hai.

7. Characteristic polynomial $\chi (s)$ — "woh equation jise poles solve karte hain"

Guess kiye bina eigenvalues dhundne ke liye, poochho: kin $\lambda$ ke liye $Av=\lambda v$ ka nonzero solution exist karta hai? Yeh exactly tab hota hai jab matrix $\lambda I-A$ kisi direction ko zero par squash karta hai, yaani uska determinant vanish ho jaata hai.

Yahan $I$ identity matrix hai (diagonal par 1's, baaki jagah 0's — "do-nothing" transformer), aur $\det$ determinant hai, ek single number jo measure karta hai ki matrix area/volume kitna stretch karta hai; zero determinant ka matlab hai yeh ek direction collapse kar deta hai.

Parent in mein se do use karta hai:

• $\chi (s)$ — system ka actual char. poly ($A$ se).
• $\alpha (s)=(s-{\mu }_1)\cdots (s-{\mu }_n)=s^n+a_{n-1}{s}^{n-1}+\cdots +a_0$ — desired wala, target poles ${\mu }_i$ se banaya gaya jo aap choose karte hain.

8. Cayley–Hamilton — "ek matrix apne khud ke polynomial ko satisfy karta hai"

"Matrix ko polynomial mein plug karna" ka matlab kya hai? $s^k$ ko matrix power $A^k$ se replace karo aur constant $c_0$ ko $c_{0}I$ se:

$\alpha (A)=A^n+a_{n-1}{A}^{n-1}+\cdots +a_{1}A+a_{0}I.$

9. Controllability matrix $\mathcal{C}$ — "kya aapka ek push har wobble tak pahunch sakta hai?"

Aapka push $u$ sirf direction $B$ ke through enter hota hai. Lekin repeated pushes dynamics ke through ripple karte hain: $B$, phir $AB$, phir $A^{2}B$, ... har step state space mein aur door tak pahunchte hain. Woh reach-vectors side by side stack karo:

Figure: $B$ aur $AB$ do arrows ki tarah. Jab woh alag-alag directions mein point karte hain (plane span karte hain, area $\ne 0$) toh aap kahin bhi pahunch sakte ho → controllable → ${\mathcal{C}}^{-1}$ exist karta hai. Jab woh parallel hain (collapse, zero area) toh ek poora direction unreachable hai → not controllable → ${\mathcal{C}}^{-1}$ nahi → Ackermann fail ho jaata hai.

10. Selector row $[0\cdots 01]$

Ek single row jo, jab kisi matrix ko multiply kare, us matrix ki bottom row pluck out kar le. Controllable canonical form mein sirf last coordinate actuated hota hai, isliye feedback gains bottom row mein jaate hain — aur yeh row unhe read off karti hai. Kuch mystical nahi: yeh "last line dekho" hai.

Pieces ko milana

Ab boxed formula mein har ek symbol ka matlab hai:

$K=\underset{\text{bottom row read karo}}{\underbrace{[0\cdots 01]}}\underset{\text{return ticket}}{\underbrace{{\mathcal{C}}^{-1}}}\underset{\text{desired dynamics, }A\text{ ko diya gaya}}{\underbrace{\alpha (A)}}.$

Equipment checklist

Right-hand side cover karo aur dekho ki reveal karne se pehle har ek ka jawab de sako.