Knowing where a system should go is not the same as having the means to move it there.
A controller may know the desired state perfectly. Sensors may reveal the current state. Yet the available actuators may not influence every direction the state can move.
Controllability asks whether the available inputs can, in principle, move the system from its present internal state to the state we need.
This is a specialist control concept beneath How Control Systems Work and the natural counterpart to How Observability Works. Observability asks whether we can know the state. Controllability asks whether we can move it.
State Has Directions
Suppose a system has several state variables: position, velocity, temperature, pressure, charge or some other internal quantities.
An actuator may influence some combinations of these states strongly and others weakly or not at all.
A heater can raise temperature but cannot directly reduce humidity. A traction motor can accelerate a train along the track but cannot steer it sideways. A pump can change flow in one branch but may have little authority over another branch separated by closed valves.
Controllability is therefore not simply “do we have an actuator?” It is “does the actuator couple into every state direction that matters for the target?”
Formal Controllability and Practical Reachability
In linear control theory, controllability has a precise mathematical meaning: whether suitable inputs can move the state through the required state space over finite time.
Real systems add actuator limits, safety constraints, rate limits and forbidden regions.
A system can be controllable in the formal idealised model and still be unable to reach a particular target fast enough or safely enough in practice.
That practical question is often better described as reachability under constraints.
One Actuator Can Sometimes Control Several States
Do not assume one actuator controls only one state.
A force applied to a moving mass changes acceleration, which changes velocity, which changes position. Through dynamics, one input can influence several state variables over time.
This is why controllability must be analysed through the system dynamics rather than by counting actuators and state variables mechanically.
Some State Directions Can Be Uncontrollable
If a particular internal mode is not influenced by any available input, the controller cannot actively move that mode.
It may have to rely on natural decay, redesign the actuator architecture, add another input or avoid operating conditions where that uncontrolled mode becomes dangerous.
This is a design problem, not a tuning problem. No amount of gain adjustment can create control authority in a direction the actuator does not influence.
Controllability and Saturation Are Different
An uncontrollable state direction is structurally unreachable through the available inputs.
Saturation is different: the input does influence the state, but the available magnitude or rate is insufficient under the present condition.
See How Control Saturation Works. The distinction matters because the repairs differ: add or reposition control authority versus enlarge or protect existing authority.
Controllability Can Change With Operating State
Nonlinear systems can become easier or harder to control in different regions.
A vehicle’s steering effectiveness changes with speed. A valve has different sensitivity near closed and near fully open. Aircraft control surfaces behave differently across flight conditions.
This is one reason a single controller parameter set may not work everywhere. Later, How Gain Scheduling Works follows this changing operating envelope.
Worked Example: A Cart on a Track
A cart moves along a straight track. Its state is position and velocity. The actuator applies forward or backward force.
Because force changes acceleration, the input can change velocity; velocity changes position. Over time, one actuator can move both state variables.
If the actuator can apply only forward force and no braking or reverse action exists, the reachable state set changes. Some desired states may be technically possible only through long passive slowing or may not be reachable within the required time.
Worked Example: Train Motion
A train’s motion along the track is influenced by traction and braking, with route gradient, resistance and adhesion shaping the dynamics.
The train cannot directly command position. It commands forces that change speed, and speed changes position.
Safe railway control therefore asks whether available traction and braking can move the train through the required speed–position trajectory within the route and adhesion constraints.
The applied railway mechanisms live under How MRT Works | It’s Mathematics.
Worked Example: Building Climate
A building controller may have heating and cooling authority but no direct humidity control.
Humidity may change indirectly through temperature and ventilation, but some humidity states cannot be driven to target reliably without dehumidification or fresh-air control.
If humidity is important, the architecture may need another actuator rather than more aggressive temperature control.
A Careful Analogy: Logistics
A warehouse wants to reduce late departures. Management can move labour between picking, packing and loading, but it cannot directly change a late inbound supplier arrival.
The control analogy asks which state variables the operation can actually influence and which are external disturbances or upstream ownership problems.
If the bottleneck is carrier capacity, reallocating warehouse labour may have almost no authority over the final outcome. The organisation needs another actuator: alternate carrier, different service promise or upstream scheduling intervention.
A Careful Analogy: Education
A teacher can influence representations, practice, feedback, pacing and task selection. The teacher cannot directly command “understanding” into existence.
The controllability analogy asks whether the available instructional actions can plausibly move the learner state that matters.
If a learner’s difficulty is rooted in an earlier prerequisite, giving more advanced-topic practice may be an actuator pointed at the wrong state direction.
The analogy remains bounded: learners are autonomous adaptive humans, not state vectors under external command. Its value is to prevent the teacher from confusing effort with effective influence.
A Careful Analogy: Government and Institutions
Institutions often declare outcomes they do not directly control.
A ministry may influence regulation, funding, standards and incentives but not command every private decision. A central authority may influence demand conditions without directly controlling every price.
The control analogy helps separate desired outcome from available policy actuator. It should never be used to pretend society is a deterministic plant.
Actuator Placement Matters
Two actuators of the same power can have very different control value depending on where they couple into the system.
A valve placed upstream can influence several branches. A local heater controls one zone. A sensor–actuator pair located near the disturbance may control a mode that a distant actuator influences only weakly.
Control architecture therefore asks not only how much authority exists, but where it enters the dynamics.
Redundancy Can Increase Reachability
Multiple actuators can provide alternate paths to the same state.
If one fails, another may preserve enough control authority for degraded operation. This is not guaranteed; two actuators can share the same common-cause failure.
The broader owner is How Redundancy Works. In the control context, redundancy matters when it preserves reachable state after actuator loss.
A Controllability Diagnostic
- Define the internal state that must be moved.
- List every available actuator or decision input.
- Map which state directions each input influences.
- Test whether some important state combinations are unaffected.
- Include actuator magnitude and rate limits.
- Include safety and forbidden-state constraints.
- Test actuator failures.
- Measure whether the target is reachable within the receiver’s required time.
- If not, decide whether to add authority, move authority, change the target or redesign the process.
Observability and Controllability Form a Pair
A system can have four broad combinations:
- observable and controllable: state can be inferred and moved;
- observable but not controllable: the system can see a problem it cannot actively fix;
- controllable but poorly observable: action exists, but state knowledge is weak;
- neither: the system is largely blind and powerless with respect to that state.
This pair is one of the cleanest ways to think about control architecture before arguing about detailed tuning.
The CivDJ Rotation
- Forward: desired state → controller → available actuators → state trajectory → achieved or unreachable target.
- Backward: start from a failed target and ask whether the system ever had enough authority over the relevant state.
- Rotate: inspect control authority as designer, operator, maintainer, finance owner and receiver.
The operator may ask for another actuator. Finance may ask whether the rare condition justifies the capacity. Reliability asks what happens after one actuator fails. The receiver asks whether the essential outcome survives.
The Civilisation Lesson
Many systems fail by confusing responsibility with control.
A person or institution can be responsible for an outcome while lacking direct authority over every variable that produces it. Good system design makes that gap visible and either supplies adequate actuators, changes the coordination structure or narrows the promise.
Controllability asks the question targets often avoid: even if we know exactly where we want to go, do the actions available to us actually span a path to get there?
Continue through How Observability Works, How Control Saturation Works and the master How X Works hub. Next: gain scheduling — how control rules change when the same actuator–plant relationship behaves differently across operating states.