Intuition The one idea behind everything here
If you can find a single "height number" that is lowest at your target and only ever slides downhill as the system moves, then the system must roll to that target — and you never had to solve its equations. This whole topic is just making the words height , lowest , and downhill precise enough to be a mathematical proof.
Before you can read the parent note Lyapunov stability you must be fluent in a stack of symbols it throws at you. We build each one from nothing, in an order where every symbol is earned before it is used.
A machine (a spacecraft, a pendulum, a car) at any instant is completely described by a short list of numbers: where it is and how fast each part is moving . That list is the state .
x
x (bold x ) is a stack of numbers — a point that pins down the whole machine right now. If it takes n numbers, we write x = ( x 1 , x 2 , … , x n ) .
Picture: an arrow from the origin to a single dot in space. For n = 2 it's a dot on a flat sheet of paper.
Why the topic needs it: stability is a statement about where the dot goes over time . No dot, nothing to track.
n and the space R n
n = how many numbers are in the state. R n (read "R-n") = the set of all possible such stacks — every dot you could ever be at.
Picture: R 1 = a number line, R 2 = a flat sheet, R 3 = a room.
Why: "x ∈ R n " is just shorthand for "the state is one of the allowed dots."
Look at the red dot above: that single point is the entire machine at this instant. As time passes, the red dot slides around and traces a curve.
The state does not sit still — it moves. To talk about how fast it moves we need velocity.
x ˙
A dot on top means rate of change per second (velocity). x ˙ 1 is "how fast x 1 is changing right now."
Picture: a little arrow attached to the red dot showing which way and how fast it is about to move. This arrow is called the velocity vector .
Why: stability asks "which way does the dot drift?" — that is exactly what the velocity arrow tells you.
Now, who decides that arrow? The physics of the machine does. There is a rule: "if you are at state x , your velocity arrow is this." That rule is the function f .
f ( x )
f is a machine that eats a state and spits out its velocity arrow . The equation of motion is simply:
x ˙ = f ( x )
"My velocity is whatever f says it is at my current position."
Picture: at every dot in the sheet, draw the arrow f assigns. You get a whole field of arrows — a vector field (see figure). Release the red dot and it flows along the arrows like a leaf on a stream.
Why the topic needs it: the entire trick of Lyapunov is to read the direction of these arrows without ever computing the leaf's full journey.
The parent says the system is autonomous : f depends only on where you are, not on what time it is . The arrow field is frozen — it looks the same at 9 AM and at midnight. This is why V later has no explicit time dependence .
Definition Equilibrium and the zero vector
0
0 (bold zero) is the state stack of all zeros — one specific dot we slide to the origin. An equilibrium is a state where the velocity arrow has length zero: f ( 0 ) = 0 .
Picture: a dot where the stream is perfectly still. Drop the leaf exactly there and it never moves.
Why: stability is always about an equilibrium — the "target" the machine should return to. We shift coordinates so this target sits at the origin, which is always allowed (just subtract off where it really is).
To say "the dot stays close to the target" we need to measure how far the dot is from the origin.
∥ x ∥
∥ x ∥ (double bars) is the length of the arrow from the origin to the dot — its plain distance. For x = ( x 1 , x 2 ) it is x 1 2 + x 2 2 (Pythagoras).
Picture: a ruler laid from origin to dot. "∥ x ∥ < ε " means "the dot is inside a ball of radius ε ."
Why: the parent's ε –δ definition of stability is entirely built from these balls — a small start-ball δ staying inside a target-ball ε .
ε (epsilon) and δ (delta)
These are just two Greek letters standing for two small radii : ε = "how far I'm willing to let the dot wander" (target ball), δ = "how tightly I must aim at the start" (start ball). Stability = for any wander-limit you pick, there's an aim-tightness that guarantees it, forever .
Definition Scalar function
V ( x )
V eats a state (a stack of numbers) and returns one single number — a scalar . Think of it as a height / altitude assigned to every dot.
Picture: a landscape draped over the sheet of states — hills and valleys. V ( x ) is the altitude directly above the dot x (see figure).
Why: Lyapunov's whole method is "find a landscape shaped like a bowl, whose bottom is the target, and check the leaf always flows downhill ." V is that landscape.
The bowl shape has a name — positive definite — which we now unpack symbol by symbol.
Definition Positive definite (PD), in words
V is positive definite if: V ( 0 ) = 0 (bottom of the bowl sits exactly at the target) and V ( x ) > 0 for every other dot (everywhere else is strictly higher ground).
Picture: a genuine bowl — one lowest point, ground rising in every direction away from it.
Contrast (semi-definite): if the ground is merely ≥ 0 and stays flat along some channel, it is only positive semi-definite — a bowl with a level trough, which cannot trap a marble rolling along the trough.
To say "downhill" precisely we need the direction of steepest uphill — that is the gradient.
∇ V
∇ V (nabla-V) is the vector of partial slopes : ∇ V = ( ∂ x 1 ∂ V , … , ∂ x n ∂ V ) . Each ∂ x i ∂ V asks "if I nudge only x i , how fast does height change?"
Picture: an arrow lying on the landscape pointing straight uphill , longest where the slope is steepest.
Why: to know if the leaf goes down, compare its velocity arrow f against the uphill arrow ∇ V . That comparison is a dot product .
⊤ and the dot product a ⊤ b
⊤ tips a column stack onto its side (a row). Writing a ⊤ b is the compact way to write the dot product a 1 b 1 + a 2 b 2 + … , which measures how much two arrows point the same way .
Picture: positive if the arrows roughly agree, zero if perpendicular, negative if they oppose.
Why this tool and not another: we want ONE number saying "is the leaf's motion uphill or downhill?" The dot product of velocity f with uphill ∇ V is exactly that number — negative means moving against uphill, i.e. downhill .
V ˙ — the rate of height change along the motion
V ˙ is "how fast the altitude the leaf sees changes per second as it flows." By the chain rule (rate of a compound quantity = slope × speed, summed over each direction):
V ˙ = ∑ i ∂ x i ∂ V x ˙ i = ∇ V ⊤ f ( x )
Picture: stand on the moving leaf and watch your altimeter. V ˙ < 0 means it reads falling .
Why the chain rule and not ∂ V / ∂ t : V has no clock inside it; all its change comes through the moving dot, whose speed the dynamics f supply. Substituting x ˙ = f is the miracle: V ˙ is computable without ever solving for the trajectory.
Mnemonic The three key symbols in one breath
f = where the leaf goes , V = how high it is, V ˙ = ∇ V ⊤ f = whether it's going down . Stability = bowl V + downhill V ˙ .
The parent's quadratic bowls look like V = x ⊤ P x .
Definition Symmetric matrix
P and quadratic form
P is a square table of numbers with P = P ⊤ (mirror-symmetric across its diagonal). x ⊤ P x is a weighted sum of squares — the simplest possible bowl.
Picture: for P = I (identity) it's the perfectly round bowl x 1 2 + x 2 2 ; other P stretch or tilt it.
Why: checking such a bowl is PD reduces to checking P has all positive eigenvalues — see Positive Definite Matrices & Sylvester's Criterion and the Lyapunov equation .
dynamics f as arrow field
Every arrow means "you need the left box to understand the right box." The two boxes positive-definite bowl and downhill V ˙ feed straight into Lyapunov's theorem — that is the whole topic.
Cover the right side; can you answer each before reading on?
What does bold x mean and what picture goes with it? A stack of n numbers = one dot in R n describing the machine right now.
What is x ˙ and who determines it? The velocity arrow at the dot; the dynamics f assign it via x ˙ = f ( x ) .
What is an equilibrium in one picture? A dot where the arrow has zero length (f ( 0 ) = 0 ) — the stream is still there.
What does ∥ x ∥ measure? The straight-line distance (arrow length) from origin to the dot.
State stability using ε and δ in plain words. For any target ball ε there is a start ball δ so starting inside δ stays inside ε forever.
What kind of object is V ( x ) ? A scalar — one number (a height) for each state; a landscape over the state space.
What makes V positive definite? V ( 0 ) = 0 and V ( x ) > 0 everywhere else — a true bowl with its bottom at the target.
What does ∇ V point toward? Straight uphill on the landscape, longest where steepest.
Why is V ˙ = ∇ V ⊤ f and not ∂ V / ∂ t ? V has no clock; all change flows through the moving dot, so chain rule gives gradient dotted with velocity f .
What sign of V ˙ means the leaf is going downhill? Negative — velocity opposes the uphill gradient.