Control theory

Control theory studies how to choose control inputs for a dynamical system so that its plant output follows a reference signal, despite control disturbances and imperfect knowledge of the controlled plant. A controller turns the available information into a control signal. The design must consider both the resulting response and stability theory: a rule that corrects a small mismatch can still produce growing deviations if the interconnection is badly chosen. ^[ref]

The central distinction is informational. Open-loop control executes planned commands; feedback control uses the measured output of the actual controlled plant to update those commands. A reference signal can describe a destination, while control disturbances describe influences that the controller did not command. A model can predict where the controlled plant should be; measurement tells the controller what actually happened. The authors provide a free full-book PDF through ^[ref].

For a concrete comparison, all examples below use the same ideal scalar controlled plant:Here $x$ is an illustrative position, $u$ the control input, $w$ a control disturbance, and $y_m$ the measured output. Both $x$ and $t$ are dimensionless quantities; the numbers are chosen to expose the mechanism, with no empirical interpretation. The ordinary differential equation says that the present control input and control disturbance accumulate into $x$. The measured output equals the entire state only in this ideal example; a general plant output need not reveal the entire state.

The controlled plant is the part of a control theory model that receives the control input and produces the plant output, including the mechanism that applies the command. The controller is specified separately. In the illustrative ordinary differential equation $\dot{x}=u+w$, the controlled plant accumulates the control input and control disturbance; it does not itself choose $u$.

A plant input is a signal supplied to the controlled plant. The chosen plant input is its control input; other influences can enter as control disturbances. This use concerns a dynamical system, rather than the input of a computation.

A plant output is a quantity exposed by the controlled plant, often modeled as $y=h(x,u,t)$. In the illustrative model it is $y=x$. An output of a computation has a different meaning.

The measured output is the information about the plant output available to the controller. It may differ from the actual plant output because of measurement noise or feedback delay. In the ideal examples $y_m=y=x$, so the controller observes the actual current state exactly.

A sensor bias is a systematic offset between the measured output and the physical plant output, modeled here as $y_m=y+b$ with constant $b$. Unlike zero-mean measurement noise, a sensor bias persists when repeated readouts are averaged. A sensor bias affects the physical meaning of a zero tracking error; for a temperature readout, independent thermometer calibration can detect it.

A control disturbance is an influence on the controlled plant that is not the chosen control input. In $\dot{x}=u+w$, a constant $w=0.2$ increases $x$ by $0.2$ per unit $t$ whenever $u=0$.

A controller is the rule or mechanism that chooses a control signal. An open-loop control can depend on a precomputed schedule; a feedback control also uses a measured output and can have its own memory, as in integral action.

A feedback gain sets the strength of a controller's response to its observations. For proportional control, $u=ke$ means that a tracking error $e=0.1$ produces a control input $u=0.2$ when $k=2$.

Actuator saturation means the applied control input reaches a device limit even if the controller requests a larger or smaller command. In the heater interpretation, the applied command is bounded between $0$ and $100\%$. If integral action continues accumulating a persistent tracking error beyond that limit, its stored integral can grow without producing the requested physical correction, delaying recovery when the goal changes. The unconstrained steady-state tracking error formulas require that the necessary command can actually be applied. ^[ref]

An ideal PID controller combines proportional control, integral action, and a term proportional to the derivative of the tracking error. With $k_d=0$, it gives proportional-integral control. The integral stores past error, while the derivative responds to its current rate of change. Practical implementations must account for measurement noise, which an ideal differentiator can amplify. ^[ref].

Controllability asks whether the available control inputs can move a dynamical system between every pair of states in its declared state space. Fixing a duration $T>0$ gives the precise condition ${\mathcal{R}}_T(x_0)=X$ for every $x_0\in X$, where $X$ is the state space and ${\mathcal{R}}_T(x_0)$ its reachable set. With freely chosen duration, every pair must instead be connectable in some permitted finite time, which may depend on the pair. Always specify admissible control inputs, deadlines and any constraints along the path. ^[ref]

The surprising part is that one control input can control several state variables by acting at different times. For the double integrator, acceleration changes velocity directly, while the accumulated velocity changes position. Starting from rest, accelerate during the first half of the journey and brake during the second: the final velocity can vanish although the final position has changed. The figure is an analytic mathematical example in dimensionless quantities, with no experimental measurements.

Controllability is useful before trying to design a controller: a requested endpoint outside the reachable set cannot be attained by clever choice of feedback control. The Kalman rank criterion makes this an exact linear algebra calculation for the stated linear time-invariant system model. It concerns whole states, so attaining one plant output is not enough to establish controllability. In quantum control, the declared state space may be an orbit constrained by conservation laws; see quantum controllability and unitary orbit of a density operator.

For the linear time-invariant system $\dot{x}=Ax+Bu$, with $A\in {\mathbb{R}}^{n\times n}$ and $B\in {\mathbb{R}}^{n\times m}$, the controllability matrix has $n$ rows and $nm$ columns. The columns of $B$ are directions influenced directly by the control inputs; $AB,A^{2}B,\dots$ account for how the dynamics carry their influence into further directions. The column space of the controllability matrix is the zero-state reachable set for any $T>0$ when inputs are unrestricted. ^[ref]

For a finite-dimensional continuous-time linear time-invariant system $\dot{x}=Ax+Bu$ on ${\mathbb{R}}^n$, with constant matrices $A,B$ over the real numbers, no state constraints, and every piecewise continuous function $u:[0,T]\to {\mathbb{R}}^m$ allowed as a control input without an amplitude bound, controllability in any fixed $T>0$ is equivalent to the controllability matrix having rank of a matrix $n$. The same equivalence holds with all square-integrable functions as inputs. A stronger useful statement iswhere $e^{AT}$ is a matrix exponential and $\mathrm{im}\mathcal{C}$ is the column space. Thus a rank of a matrix $r<n$ leaves only an $r$-dimensional affine subspace reachable at that time. ^[ref]

The Kalman rank criterion allows arbitrarily large commands as $T$ shrinks. It does not assert controllability with bounded control inputs, nor test a general nonlinear dynamical system just by counting its inputs. For the double integrator, $B=(0,1{)}^T$ and $AB=(1,0{)}^T$ give two independent directions, despite there being only one control input.

For $T>0$ and the constant matrices over the real numbers in the Kalman rank criterion, the finite-time Controllability Gramian accumulates the influence of all command times. Its column space equals that of the controllability matrix, and $W_T$ is a symmetric positive-definite matrix exactly when the controllability matrix has rank of a matrix $n$. If so, put $\delta =x_1-e^{AT}{x}_0$. Among square-integrable functions steering $x_0$ to $x_1$ at $T$, the commandminimizes ${\int }_0^T\Vert u(t){\Vert }^{2}dt$, with value ${\delta }^T{W}_T^{-1}\delta$. Here $T$ in a matrix superscript denotes transpose. This cost is a squared input norm, not necessarily physical energy, and the optimizing command need not satisfy an amplitude bound. ^[ref]

For the double integrator, $e^{As}B=(s,1{)}^T$, so direct integration givesFor a rest-to-rest distance $d$, $u_{\ast }(t)=6d/T^2-12dt/T^3$ and the minimum squared input norm is $12d^2/T^3$. Compare the two constant commands in the double integrator example, whose cost is $16d^2/T^3$: both attain the target, but the smooth command costs less. At $T=2$, $d=1$, however, $u_{\ast }$ peaks at $3/2$, so it violates $|u|\le 1$ while the two constant commands remain admissible. Controllability, optimal control, and constrained reachability in control theory answer different questions.

Quantum controllability concerns the transformations reachable under specified quantum control inputs. Unitary operator controllability concerns entire propagators; density operator controllability concerns conjugation on every spectral orbit; pure-state controllability concerns normalized state vectors or their physical rays. These notions need not coincide.

A finite-dimensional closed quantum control system with fixed drift and control Hermitian matrices, real control inputs, and dynamics linear in the state at fixed controls and affine in the controls at fixed state. Its Dynamical Lie algebra is generated over the reals by $i{H}_0,i{H}_1,\dots$. Reachability depends on the admissible controls, allowed duration and phase convention.

The dynamical Lie algebra of a finite-dimensional Hamiltonian control system is the real span of its skew-Hermitian drift and control generators and all their nested commutators. The sign choice $iH$ or $-iH$ gives the same algebra. Its connected group determines reachable propagators under the usual freely timed, real-control assumptions. An operator commuting with every generator also commutes with this algebra, imposing invariant sectors.

Let $H_0$ be diagonal and $H_1={\sum }_{j=1}^{N-1}{d}_j(E_{j,j+1}+E_{j+1,j})$, with every $d_j\ne 0$ and the adjacent absolute energy gaps distinct and nonzero. The squared drift adjoint action has distinct eigenvalues on these coupling components, so Lagrange interpolation isolates every adjacent symmetric coupling. One more commutator gives its antisymmetric partner. Their mutual brackets produce adjacent traceless diagonals, while brackets along the connected chain give all longer-range matrix units. Thus the special unitary Lie algebra is generated. Any nonzero-trace generator also supplies the central direction, giving all of $\mathfrak{u}(N)$. This is a sufficient condition in the finite-dimensional freely timed real-control model; connectivity alone does not remove frequency degeneracies.

A drift Quantum Hamiltonian is strongly regular when its energy levels are nondegenerate and distinct unordered transitions have distinct absolute nonzero Bohr frequencies. Sometimes regularity is required only on the graph of allowed control transitions. This lets polynomials in the squared drift adjoint action separate different coupling quadratures and permits frequency-selective resonant quantum control. This notion is unrelated to a strongly regular graph.

For $N\ge 2$, let $\mathfrak{g}$ be a Lie subalgebra of $\mathfrak{u}(N)$ with codimension one. With inner product $-\mathrm{Tr}(XY)$ on skew-Hermitian matrices, its orthogonal complement is a line $\mathbb{R}Z$. Invariance of the inner product shows that $[X,Z]$ lies in that line for every $X\in \mathfrak{g}$; skew-adjointness of the adjoint action makes $[X,Z]$ orthogonal to $Z$, hence zero. A nonscalar skew-Hermitian $Z$ has a centralizer of dimension at most $(N-1{)}^2+1<N^2-1$, by its orthogonal eigenspace decomposition, and cannot centralize all of $\mathfrak{g}$. Thus $Z$ is scalar and $\mathfrak{g}=\mathfrak{su}(N)$. This gives density operator controllability and all determinant-one gates, but not the central phase direction needed for unitary operator controllability.

For a nonsingular skew-symmetric unitary $J$, if each skew-Hermitian control generator obeys $X^{T}J+JX=0$, then so do their real linear combinations and commutators. Differentiating $U^{T}JU$ shows that it stays equal to $J$. The reachable group is therefore contained in a unitary conjugate of the compact symplectic group. This is preservation of an alternating bilinear form, not commutation with $J$.

For $U^{T}JU=J$ with real orthogonal skew-symmetric $J$, the dual $\tilde{\rho }=J{\rho }^T{J}^{†}$ transforms as $\widetilde{U\rho U^{†}}=U\tilde{\rho }{U}^{†}$. Hence $I_J(\rho )$ is constant on the reachable unitary orbit of a density operator. Different values can obstruct density operator controllability even when the ordinary spectra agree. For a four-dimensional $J$ pairing indices $1,4$ and $2,3$, the two diagonal arrangements $(a,a,b,b)$ and $(a,b,b,a)$ have invariants $4ab$ and $2(a^2+b^2)$.

Let $H_0=\mathrm{diag}(-{\omega }_1,0,-{\omega }_2)$ and $H_1=d_1(E_{12}+E_{21})+d_2(E_{23}+E_{32})$, with nonzero $d_j$, distinct positive ${\omega }_j$. The squared adjoint action of $-i{H}_0$ has different eigenvalues $-{\omega }_1^2,-{\omega }_2^2$ on the two adjacent skew-Hermitian couplings. Linear combinations of $-i{H}_1$ and its double commutator with $-i{H}_0$ isolate both couplings. Further commutators give their real antisymmetric partners, two traceless diagonal generators, and both $1$–$3$ generators, hence the special unitary Lie algebra $\mathfrak{su}(3)$. The drift has nonzero trace; subtracting its traceless part supplies $iI$, giving the Lie algebra of the unitary group $\mathfrak{u}(3)$ and unitary operator controllability under the usual freely timed real-control assumptions.

Let $J$ be a symmetric unitary matrix and every skew-Hermitian control generator obey $x^{T}J+Jx=0$. The condition is closed under real linear combinations and commutators, and evolution preserves $U^{T}JU=J$. In a unitary basis where $J$ becomes the identity, the generators are real antisymmetric, so the Dynamical Lie algebra is contained in $\mathfrak{so}(N)$. The state quantity $|{\psi }^{T}J\psi |$ is invariant under the reachable evolution and insensitive to global phase, obstructing unrestricted pure-state controllability. This is a bilinear-form symmetry; it does not assert that $J$ commutes with the Hamiltonians.

For $N\ge 2$, suppose a skew-Hermitian Lie algebra becomes all of $\mathfrak{u}(N)$ after adding the scalar identity direction. Scalar matrices commute, so $[\mathfrak{g},\mathfrak{g}]=[\mathfrak{u}(N),\mathfrak{u}(N)]=\mathfrak{su}(N)$ is contained in $\mathfrak{g}$. If $\mathfrak{g}$ contains any nonzero-trace generator, subtracting its traceless part supplies $iI$, and $\mathfrak{g}$ was already all of $\mathfrak{u}(N)$. This permits using identity-augmented commutator calculations without assuming an additional physical identity control.

For freely chosen control duration in a finite-dimensional closed system, pure-state controllability means the reachable group acts transitively on normalized states or their rays. Up to unitary conjugation, the possible Dynamical Lie algebras are $\mathfrak{u}(N)$, $\mathfrak{su}(N)$, and for even $N=2n$ the standard compact symplectic Lie algebra $\mathfrak{sp}(n)$, with or without $\mathbb{R}iI$. For $N\ge 4$, the symplectic possibilities are pure-state controllable but generally lack density operator controllability.

A closed system has density operator controllability when every pair of density operators with the same spectrum can be connected by its controls. Under the standard finite-dimensional Hamiltonian assumptions, this holds exactly for Dynamical Lie algebra $\mathfrak{u}(N)$ or $\mathfrak{su}(N)$. This definition ranges over every spectral orbit; a particular degenerate spectrum can be controllable under a smaller group.

A finite-dimensional closed-system control model is unitarily operator controllable when its reachable propagators include every unitary operator. For the usual real Hamiltonian controls and freely chosen duration, this requires the Dynamical Lie algebra to be $\mathfrak{u}(N)$. If global phase is ignored, $\mathfrak{su}(N)$ also suffices; exact access to all of $U(N)$ still requires the central phase direction.

Reachability in control theory asks about a particular initial state $x_0$ and target $x_1$: does some admissible control input make $x(T)=x_1$ at the declared time $T$, or at some allowed duration? Controllability requires this for every pair in the declared state space. This is a property of state evolution, distinct from graph reachability.

For a linear time-invariant system, solving the ordinary differential equation by a matrix exponential givesThe first term is the unforced endpoint. Reachability in control theory asks whether the control input can supply exactly the displacement $x_1-e^{AT}{x}_0$. The timing matters because the same command is carried by different amounts of subsequent dynamics. ^[ref]

Here is a contrasting mathematical example with one shared control input:The controllability matrix has rank of a matrix one. Since $\frac{d}{dt}(z_1-z_2)=0$, the difference is invariant under every allowed control input. From $(0,0)$ the reachable set is the diagonal $z_1=z_2$ when input amplitudes are unrestricted: $(1,1)$ is reachable at $T=2$ using $u=1/2$, while $(1,0)$ is impossible at every duration. Having a control input appear in both state equations did not give independent control of both states. This reachable set is pictured alongside the bounded double integrator below.

The reachable set ${\mathcal{R}}_T(x_0)$ collects all endpoints attainable from $x_0$ at exactly $T$ by admissible control inputs. If arrival any time up to $T$ is allowed, use ${\mathcal{R}}_{\le T}(x_0)={\bigcup }_{0\le t\le T}{\mathcal{R}}_t(x_0)$, with ${\mathcal{R}}_0(x_0)=\{x_0\}$. These definitions differ: a general dynamical system may drift away after early arrival, so waiting need not preserve the target. The reachable set is always relative to a declared input class and any constraints on the entire path.

For the double integrator starting from $(0,0)$, now require measurable functions as control inputs with $|u(t)|\le 1$ almost everywhere. Write $(p,v)$ for its endpoint at $T>0$. Two integrals determine the endpoint:Our mathematical example has the exact reachable setTo see the upper boundary, fix $v$: put $u=1$ first and $u=-1$ last, switching at $s=(T+v)/2$. This maximizes the weighted integral for position because early commands have the largest weight $T-t$; the required net velocity fixes $s$. Formally, $(T-t)-(T-s)=s-t$ has the same sign as this maximizing command, so changing to any other admissible command with the same integral cannot increase $p$. Reversing the order gives the lower boundary. Pointwise convex combinations of these two commands preserve $v$ and attain every intermediate $p$, so the formula describes the whole region, not merely a bound. The extreme commands are bang-bang control.

In particular, the bounded double integrator can move from rest to position $d$ and rest again exactly when $|d|\le T^2/4$. Its minimum duration is therefore $2\sqrt{|d|}$ in these normalized units. With freely chosen time every rest-to-rest distance is attainable; with a fixed deadline many are excluded. Changing the amplitude bound to $|u|\le a$ with $a>0$ scales this condition to $|d|\le a{T}^2/4$. This shows why specifying the controls matters even when the Kalman rank criterion gives full rank of a matrix.

In closed quantum control, a density operator reachable set lies inside its unitary orbit of a density operator, since coherent evolution preserves the spectrum. That orbit need not be fully reachable under the available control inputs.

The double integrator is the linear time-invariant system in which one control input is integrated into velocity, and velocity into position. It is a mathematical idealization of one-dimensional motion under commanded acceleration; for a constant mass $m$ and net force $F$, Newton's second law gives $u=F/m$. The worked examples here use normalized dimensionless quantities and assume no control disturbances, rather than describing a laboratory record. ^[ref]

Writing $x=(p,v{)}^T$ givesThe Kalman rank criterion therefore gives controllability in every $T>0$ for unrestricted real control inputs. More concretely, choose constants $a,b$ on the first and second halves. Starting at $x_0=(p_0,v_0{)}^T$, direct integration givesThe endpoint map has determinant $T^3/8\ne 0$, so these two amplitudes can supply any desired endpoint displacement. For a rest-to-rest move of distance $d$, take $a=4d/T^2$, $b=-4d/T^2$. At $T=2$ and $d=1$ these are the two commands in the first figure. No impulse is needed; shorter deadlines require larger finite commands.

An open-loop control is prescribed as a function of time, possibly after calculation from an initial state and a dynamical model. It is executed without using observations of the actual current state to change the input. A simulated feedback control law can generate an open-loop waveform, but the resulting laboratory control does not automatically correct errors in the actual state or model.

For the ideal controlled plant with $w=0$, plan the control inputIntegration gives $x(t)=\min (t,1)$, so the planned command reaches the reference signal $r=1$ at $t=1$. If an unplanned constant control disturbance $w=0.2$ begins at $t=2$, the same open-loop control gives $x(t)=1+0.2(t-2)$ for $t\ge 2$: $x(4)=1.4$ and $x(8)=2.2$. Knowing the nominal response did not provide a way to correct the actual plant output.

Quantum control chooses externally adjustable interactions to steer a quantum state or implement a desired unitary operator. Its achievable transformations depend on the Hamiltonian or Lindblad equation, the available control inputs, and physical constraints such as bandwidth and amplitude. Quantum optimal control optimizes performance within those constraints.

Individual quantum control of spins by exploiting different resonance frequencies. A sufficiently narrow-band radio frequency pulse can predominantly affect one resonance when the separation is large compared with the excitation bandwidth and coupling effects are accounted for. Equal resonance frequencies remove this frequency-based distinction. Whether individual gates remain possible then depends on other available interactions, spatial resolution, or differences in control coupling; frequency equality alone is not a universal impossibility theorem.

Let $S$ be the SWAP gate. If the drift Hamiltonian operator and every allowed control Hamiltonian operator commute with $S$, every reachable time-evolution operator also commutes with $S$. Indeed, $SUS$ satisfies the same initial-value Schrödinger equation as $U$ and starts at $I$, so uniqueness gives $SUS=U$. Yet $S(Z\otimes I)S=I\otimes Z\ne Z\otimes I$. An individual Pauli Z gate is consequently unavailable within that symmetric control model. Equal nuclear resonances, isotropic exchange interaction, and identical coupling to a common homogeneous radio frequency field provide an example. An added interaction that breaks this symmetry can remove the obstruction.

For $H=J(XX+YY+ZZ)/2$, $U=e^{-itH}$, and an individually applied Pauli Z gate $P=Z\otimes I$, conjugation gives $PHP=J(-XX-YY+ZZ)/2$. The paired Pauli operators $XX,YY,ZZ$ commute with one another, soTwo equal free-evolution intervals, each of duration $t$, therefore cancel exchange while retaining the Ising coupling. This spin echo is exact for instantaneous ideal gates; a finite pulse requires accounting for the interaction during the pulse. A simultaneous Pauli Z gate on each spin commutes with $H$ and cannot perform the same cancellation. The physical availability of the individual gate is a separate quantum control requirement.

On a connected transition chain, independently selected resonant pulses realize embedded two-level rotations with tunable angle and phase. Eliminate target amplitudes from one end of the chain using complex Givens rotations, and invert those rotations to prepare the target from a basis state. The pulse angles come from amplitude ratios and the phases from relative complex arguments. Spectator components are unchanged by each embedded rotation. Work in the interaction picture and compensate the known drift phases at the chosen final time when the target is specified in the laboratory frame.

An Ising coupling generates $e^{-i\varphi Z\otimes Z}$ with $\varphi =\int J(t)dt$ in units $\hslash =1$. It is entangling precisely when $\varphi$ is not an integer multiple of $\pi /2$. A $\pi$ rotation on one qubit reverses its sign, permitting an ideal echo to cancel an always-on coupling.

A $\pi$ rotation about the x-axis on one qubit conjugates $Z\otimes Z$ to its negative. With ideal instantaneous pulses, two equal free-evolution intervals separated by this pulse and its inverse cancel an Ising coupling exactly. Finite-duration pulses require accounting for the coupling during the pulses; the error is small only when $|J|$ times the pulse duration is small, or when the control is explicitly designed to compensate it.

In coherent quantum feedback, a quantum controller interacts with the controlled system without first converting the control signal into a classical measurement record. Optical feedback through a beam splitter is one example. The input-output interconnection must preserve the physical quantum dynamics, including noise and commutation relations; a classical state-feedback formula does not automatically define a valid quantum controller.

The rotating-wave approximation to a weak quadrature coupling of a cavity mode to a broadband field retains photon-exchange terms and removes counter-rotating pair creation and annihilation. Flat coupling around resonance produces the displayed interaction in units $\hslash =1$. Its short-time propagator is generated by $\sqrt{\gamma }(ad{B}^{†}-a^{†}dB)$; the Gaussian quantum noise Ito table supplies the quadratic drift correction. The resulting passive cavity has the passive cavity input-output transfer function.

For a broadband bosonic input with $[b(t),b^{†}(s)]=\delta (t-s)$, disjoint field increments commute and $[d{B}_t,d{B}_t^{†}]=dt$. An adapted system operator commutes with the next incoming increment. Noise increments scale as $\sqrt{dt}$, so quadratic products contribute to first-order stochastic evolution.

For a zero-mean Gaussian bosonic input, $d{B}^{†}d{B}^{†}=M^{\ast }dt$ completes the displayed table, with $N\ge 0$ and $|M{|}^2\le N(N+1)$. The vacuum has $N=M=0$; an unsqueezed thermal input has $M=0$. Products with $dt$ and higher differentials vanish in the first-order Ito equation. The difference between the first two products preserves the bosonic commutator.

For $\dot{a}=-\gamma a/2-\sqrt{\gamma }{b}_0$ and $b_1=\sqrt{\gamma }a+b_0$, zero-initial-data Laplace transforms give the displayed transfer function. For $\gamma >0$ its pole is in the open left half-plane, while $|G(i\omega )|=1$ and its zero lies in the right half-plane. Passing through $-1$ in this plant's frequency plot is not evidence of an unstable open cavity: a Nyquist stability criterion needs an explicitly specified feedback return ratio.

With $b_0=\beta b_{\mathrm{in}}-\alpha b_1$ and $b_2=\alpha b_{\mathrm{in}}+\beta b_1$, real ${\alpha }^2+{\beta }^2=1$ and $|\alpha |<1$, eliminate the internal fields to obtain $\dot{a}=-{\gamma }_{\mathrm{eff}}a/2-c{b}_{\mathrm{in}}$, where $c=\beta \sqrt{\gamma }/(1+\alpha )$ and $c^2={\gamma }_{\mathrm{eff}}$. The internal output transfer is $b_1/b_{\mathrm{in}}=\beta G/(1+\alpha G)$; the external output transfer is $b_2/b_{\mathrm{in}}=(\alpha +G)/(1+\alpha G)=(s-{\gamma }_{\mathrm{eff}}/2)/(s+{\gamma }_{\mathrm{eff}}/2)$. The loop is asymptotically stable in this nondegenerate range. At $\alpha =1$ the cavity mode is undamped and decoupled; at $\alpha =-1$ the instantaneous algebraic feedback connection is singular.

Design Lindblad operators and a Quantum Hamiltonian so that desired density operators are stationary and unwanted perturbations decay. A dissipative interaction can change the spectrum and purity of a density operator, reaching targets unavailable to purely coherent Hamiltonian engineering. The engineered stationary state must be physical, and uniqueness alone must be distinguished from an explicit proof of asymptotic convergence.

Hamiltonian engineering designs accessible interactions or their effective time average to implement desired quantum states or unitary operators. The available control inputs, Quantum Hamiltonian generators and laboratory constraints determine which transformations can actually be achieved. Model-based open-loop control computes a waveform before running it; adaptive experiments instead evaluate its performance on the physical system.

Such a quantum state remains an eigenstate of $H_0+f(t)H_1$ for every control input. Its projector commutes with all available generators, so its population is conserved. This obstructs full-space pure-state controllability even when the orthogonal bright subspace can be controlled. An instantaneous dark state of a slowly varying Quantum Hamiltonian instead supports adiabatic quantum control if a spectral gap and slowness condition are maintained.

Fast prescribed pulses change the frame in which the system evolves during successive intervals. To leading order over a short control cycle, the effective Quantum Hamiltonian is the duration-weighted mean of these conjugated generators. The Baker--Campbell--Hausdorff formula supplies commutator corrections. Symmetrizing the cycle can cancel unwanted interactions or retain a desired one. The approximation requires control cycles short compared with the relevant dynamical timescales and a suitable account of finite pulse duration.

Design a slowly varying Quantum Hamiltonian path whose relevant initial eigenstate connects to a desired final eigenstate. Under the gap and slowness hypotheses of the quantum adiabatic theorem, the state follows that instantaneous eigenstate up to phase. The path, spectral gap and duration all matter; a level crossing or excessive speed can spoil the intended state transfer.

For a resonant three-level chain with couplings ${\Omega }_1,{\Omega }_2$, the zero-energy dark state of a driven Hamiltonian is $|D⟩=({\Omega }_2|1⟩-{\Omega }_1|3⟩)/\sqrt{{\Omega }_1^2+{\Omega }_2^2}$. Turning on the target-to-intermediate coupling first and sweeping ${\Omega }_1/{\Omega }_2$ from zero to infinity rotates this state from $|1⟩$ to $-|3⟩$. The other eigenvalues are $\pm \sqrt{{\Omega }_1^2+{\Omega }_2^2}$, so $\hslash |\dot{\theta }|\ll \sqrt{{\Omega }_1^2+{\Omega }_2^2}$ controls nonadiabatic leakage. Ideally the intermediate state remains unoccupied, suppressing spontaneous-emission loss. The gap must not close while the dark state is rotating.

A near-resonant drive selects a transition and, in an appropriate rotating frame and rotating-wave approximation, controls a two-level effective Quantum Hamiltonian. If $H_{\mathrm{eff}}=\Omega (t){\sigma }_x/2$ with fixed axis, then $U(T)=\exp [-i{\sigma }_x{\int }_0^T\Omega (t)dt/2]$. Pulse area fixes the rotation angle; drive phase chooses a transverse axis. Selectivity and neglect of rapidly oscillating terms require the appropriate detuning and drive-strength regime.

For a fixed dimensionless Hermitian operator $A$ driven by $H(t)=\Omega (t)A$, all instantaneous generators commute. The exact unitary operator is $U(T)=\exp (-i\theta A)$, so the area, rather than the detailed envelope, fixes the rotation. For $A^2=I$, it is $\cos \theta I-i\sin \theta A$. This convention uses the Quantum Hamiltonian coefficient directly; a convention with $\hslash {\Omega }_{R}A/2$ instead has $\theta =\frac{1}{2}\int {\Omega }_{R}dt$.

For a Hermitian operator satisfying $A^3=A$, its squared operator $P=A^2$ is the projector onto the active eigenvalue-$\pm 1$ subspace. All remaining states have eigenvalue zero. Separate even and odd powers in the matrix exponential to obtain the displayed rotation on the active subspace and the identity on its spectator subspace. Thus an embedded two-level coupling need not square to the full Hilbert-space identity.

For $\dot{\rho }=-i[H_0+f{H}_1,\rho ]$ and ${\dot{\rho }}_d=-i[H_0,{\rho }_d]$, the Hilbert-Schmidt distance gives $V=\Vert \rho -{\rho }_d{\Vert }^2/2$. The choice $f=\mathrm{Tr}({\rho }_d[-i{H}_1,\rho ])$ yields $\dot{V}=-f^2$. Decrease alone does not prove convergence: orthogonal pure eigenstates of a diagonal $H_0$ give zero input for a transverse Pauli control even though their distance remains nonzero. Such a law can be simulated to produce an open-loop control.

For $V=\Vert \rho -{\rho }_d{\Vert }_{\mathrm{HS}}^2$, with both states evolving under the same drift and only $\rho$ receiving $f{H}_1$, cyclicity of the trace cancels the drift terms and gives the displayed identity. The scalar $C$ is real because $-i[H_1,\rho ]$ is Hermitian. Choosing $f=kC$, $k>0$, makes $\dot{V}=-2k{C}^2\le 0$. This is nonincrease, not by itself convergence to the target: additional invariant-set and reachability conditions are needed.

Take $H_0=0$, $H_1={\sigma }_y/2$, ${\rho }_d=(I+{\sigma }_z)/2$ and $\rho =(I+{\sigma }_x)/2$. The overlap is $f=1/2$, but $C=\mathrm{Tr}({\rho }_d[-i{H}_1,\rho ])=-1/2$, so the Hilbert-Schmidt Lyapunov feedback identity gives $\dot{V}=1/2>0$. A target overlap and the derivative of that overlap along the control direction are different quantities.

Feedback control selects a control input using observations of the actual controlled plant. In a tracking loop, compare the reference signal $r$ with the measured output $y_m$ to form the tracking error $e=r-y_m$, then let the controller change $u$. The control input changes the plant output, which is measured again: that return path closes the loop. A controller can use earlier observations as well as the latest one. ^[ref]

For the ideal controlled plant, proportional control $u=2(1-x)$ initially commands $u(0)=2$. As $x$ approaches $1$, the command shrinks. When a positive control disturbance pushes $x$ above $1$, the tracking error becomes negative and the control input pushes back. The result is a bounded offset rather than the continuing drift of open-loop control. Integral action can remove that offset by remembering past tracking errors.

A control selected from the observed current state, rather than specified only as a function of the initial state and time. For $x_{t+1}=x_t+u_t$ and stage cost $x_t^2+2x_t{u}_t+7u_t^2$, the infinite-horizon feedback $u_t=-x_t/3$ gives $x_t=(2/3{)}^t{x}_0$ and total cost $2x_0^2$. The finite-horizon Bellman coefficients satisfy ${\Pi }_0=0$, ${\Pi }_s=6(1+{\Pi }_{s-1})/(7+{\Pi }_{s-1})$ and increase to two, proving optimality.

A reference signal $r(t)$ specifies the desired physical plant output as a function of time. A fixed reference signal is often called a setpoint; a changing reference signal describes a trajectory to follow. A reference signal is a goal, whereas the control input $u(t)$ is the command chosen to achieve that goal. ^[ref]

Asking for a temperature of $40{}^{\circ }\mathrm{C}$ does not specify a heater command of $40\%$: dimensional analysis distinguishes the units of these quantities. The reference signal and the measured output must describe the same quantity in the same coordinates before their difference can define a tracking error. An unconverted sensor readout cannot be subtracted directly from a desired temperature.

For example, let $y=T-T_0$ be temperature rise above a fixed ambient baseline $T_0$, and write the desired temperature as $T_d=T_0+r$. A Heaviside step function gives $r(t)=R$ for $t\ge 0$, with $r(t)=0$ before the change: the target rise becomes $R$ at once, although a physical controlled plant generally cannot jump at once. A ramp reference signal $r(t)=vt$ instead asks the temperature rise to increase at a fixed rate $v$. If $R=1{}^{\circ }\mathrm{C}$ and $v=0.2{}^{\circ }\mathrm{C}{\mathrm{s}}^{-1}$, the first reference signal asks for a lasting one-degree rise; the second asks for a one-degree rise at $t=5\mathrm{s}$ and a two-degree rise at $t=10\mathrm{s}$. Here degree Celsius differences are also kelvin differences. A target that grows forever is useful for analytical tests but eventually exceeds a real heater's range.

To make the distinction between a goal and a command concrete, the examples in steady-state tracking error use the first-order low-pass filter as a new controlled plant, rather than the accumulating controlled plant $\dot{x}=u+w$ used in the earlier feedback demonstration:The time constant $\tau$ has units of time, $K$ has units of plant output per control input, and the additive control disturbance $d$ has the units of $y$. This $d$ is an equivalent output offset, not the rate-like $w$ in $\dot{x}=u+w$. A heater interpretation uses $y$ in degree Celsius differences and $u$ in percentage points of heater command, with $K$ in degrees per percentage point; the control input is a command, not a direct measurement of delivered power. With exact measurement, holding $y=R$ requires $u=(R-d)/K$. Thus even a constant reference signal can require a persistent nonzero control input. The numerical model below is deliberately synthetic and has no fitted laboratory parameters.

A ramp reference signal changes at a constant rate $v$ from $t=0$ onward. Its derivative is $v$ for $t>0$. A ramp reference signal tests whether a controller can keep following a moving goal, rather than merely reaching a fixed one; a finite steady-state tracking error can coexist with an unbounded reference signal.

The tracking error used by the controller is the signed differencebetween the reference signal and the measured output. A positive tracking error means the readout is below the desired value; a negative tracking error means it is above. With exact measurement $y_m=y$, $r=1$ and $y=1.1$ give $e=-0.1$. For a positive-response controlled plant, positive feedback gain in proportional control $u=k_{p}e$ then reduces the plant output. The sign of the controller must also account for the sign of the controlled plant's response: a positive command that reduces the plant output requires a different sign choice. Kantor uses the opposite tracking error convention, $PV-SP$, together with a minus sign in his controller; translating both signs gives the convention here. ^[ref]

Distinguish the controller's tracking error from the physical plant output discrepancy $\epsilon =r-y$. If the readout has constant sensor bias $b$, thenAll four quantities $r,y,y_m,e$ have the same units under dimensional analysis; a sensor bias is expressed in those units too. A readout that is too high by $0.2{}^{\circ }\mathrm{C}$ can make $e=0$ when the physical temperature is $0.2{}^{\circ }\mathrm{C}$ below the goal. Integral action integrates the tracking error it receives; it cannot infer this missing calibration information from that tracking error alone. Thermometer calibration against an independent reference addresses a different problem from tuning a feedback gain.

A motivating apparatus/model comparison is Park, Martin, Kelly and Hedengren's temperature-control laboratory: two TIP31C heaters with attached TMP36GZ sensors and heat sinks, an Arduino, a power supply, and a USB-connected computer. The readout converts sensor voltage to temperature; software records that measured output while specifying heater commands from $0$ to $100\%$. For the single-heater test, one command changes in steps and the other heater stays off. Parameters are fitted to one run, then a different command sequence tests predictions without feeding measured temperatures into the simulation. The authors observe that their richer thermal model predicts the response better than the approximation with a single time constant and a delayed control input, especially after command changes. The real sensor has its own thermal lag. They also repeat the same command profile at different ambient temperatures and find changes in the response, motivating ambient-condition reporting. This experiment measures a command-to-readout response, not independently the exact heater temperature or delivered power. ^[ref]