Sometimes the controller knows the correction it wants, but there is more than one way to produce it.
An aircraft can create a required moment using several control surfaces. A vehicle can split braking between regenerative and friction systems. A ship can combine rudders and thrusters. A building can meet a cooling demand through several chillers. The high-level objective is one thing; the available actuators are many.
Control allocation is the mechanism that distributes a desired system-level control action across multiple actuators.
This is the final specialist branch in the current Feedback & Control batch. It sits beneath How Control Systems Work and connects directly to How Controllability Works and How Control Saturation Works.
Separate the Desired Effect From the Device That Produces It
A high-level controller can think in terms of the effect the system needs: braking force, roll moment, cooling capacity, thrust vector or some other aggregate correction.
The allocator then asks: which combination of available actuators should produce that effect now?
This separation is useful because it keeps system-level control logic independent from some details of actuator configuration. An actuator can fail or become saturated and the allocator can redistribute the requested effect if enough alternate authority remains.
Actuators Have Different Effectiveness
Two actuators rarely produce identical effects.
One may be stronger in one direction. Another may have less delay. One may consume more energy. Another may create unwanted secondary motion. Effectiveness can also change with operating condition.
Control allocation therefore needs a model linking actuator commands to the system-level effect they produce.
In many engineering systems this relationship is represented by an actuator-effectiveness matrix or an equivalent model. The allocator solves the inverse problem: choose actuator commands that best realise the requested effect.
There May Be Many Valid Solutions
If a system has more actuator freedom than the minimum needed to create the requested control effect, several actuator combinations may work.
The allocator can then optimise a secondary objective:
- minimum energy use;
- minimum actuator wear;
- maximum remaining control margin;
- minimum noise or vibration;
- maximum use of regenerative energy;
- balanced thermal load;
- priority preservation for another future manoeuvre.
This is why control allocation is more than dividing a command evenly.
Constraints Shape the Allocation
Every actuator has limits.
- maximum and minimum command;
- maximum rate of change;
- temperature or duty-cycle limits;
- forbidden combinations;
- fault states;
- operating regions where effectiveness falls.
The allocator must choose a feasible combination inside these constraints.
If the requested effect exceeds the combined feasible authority, allocation cannot manufacture missing control. The higher-level system must accept degraded performance, change the target or enter another mode.
Allocation Preserves Margin
A naive allocator can use one actuator almost to saturation while leaving others lightly loaded.
That may meet the present command but leave little authority for the next disturbance.
A better allocator may spread effort to preserve control margin. This is especially valuable when the system needs to respond in several axes or when future demands are uncertain.
Good allocation therefore considers not only current success, but the shape of the remaining reachable state.
Failure Reallocation
One of the strongest reasons to separate control demand from actuator assignment is fault tolerance.
If one actuator fails, the system updates the available set and reallocates the desired correction across the survivors.
This only works if the remaining actuators still span enough control authority. Redundancy must be functionally useful, not merely numerical.
See How Redundancy Works for the wider mechanism of independent alternatives and common-cause risk.
Worked Example: Blended Braking
A rail vehicle or electric vehicle may have regenerative braking and friction braking available.
The high-level controller requests a total deceleration or braking force. The allocation layer decides how much each braking system contributes.
Regenerative braking may be preferred because it recovers energy and reduces friction wear, but it has limits depending on speed, traction conditions and the electrical system’s ability to accept power. Friction braking supplies the remainder and remains essential when regenerative authority is insufficient.
The system-level target is stable while the actuator mix changes continuously.
Worked Example: Aircraft Control
An aircraft may have several control surfaces capable of contributing to roll, pitch or yaw moments.
A high-level controller requests the required moment. The allocator chooses surface deflections that produce it while respecting limits and perhaps reducing drag or avoiding overuse of one surface.
If one surface becomes unavailable, the allocator can attempt to redistribute the moment across remaining surfaces.
The success of reallocation depends on controllability after the fault.
Worked Example: Chiller Plant
A building requires a certain cooling output. Several chillers can provide it.
The allocator can decide which machines run and at what load, considering efficiency curves, minimum stable load, maintenance state and reserve margin.
Equal sharing may not be optimal. Running one machine near its efficient region and keeping another available as reserve can be better than splitting demand 50–50.
A Careful Analogy: Logistics Labour
A warehouse needs to recover two hours of backlog before carrier cut-off. Several “actuators” exist in an operational sense: reassign trained staff, open overtime, change wave priority, use another packing line or move work to another site.
The allocation analogy asks which combination meets the required recovery while respecting skills, cost, safety and future capacity.
The analogy is bounded because people are not interchangeable actuators. Training, fatigue, fairness and employment rules are real constraints and moral considerations, not coefficients to optimise away.
A Careful Analogy: Education Support
A learner needs one capability repaired. Several interventions are possible: explanation, worked example, retrieval practice, visual representation, prerequisite review or peer discussion.
The control-allocation analogy asks how the teaching system combines available actions rather than applying all of them at maximum strength.
The learner’s response determines the next allocation. Again, this is an analogy for intervention choice, not a claim that human learning is a linear actuator system.
A Careful Analogy: Public Services
A surge in public demand can be met through several mechanisms: temporary staff, digital self-service, appointment extension, triage, cross-agency support or demand smoothing.
The useful allocation question is: which combination preserves the essential service with acceptable cost and burden while retaining reserve for further shocks?
The public context adds legitimacy, equity and statutory constraints that engineering optimisation alone cannot decide.
Priority Allocation
Some control objectives are more important than others.
An allocator may preserve safety-critical control first, then performance, then efficiency. When authority is scarce, lower-priority objectives are relaxed before protected invariants.
This hierarchy should be explicit before saturation occurs. Otherwise crisis forces an improvised priority order.
Allocation Can Change With Operating State
Actuator effectiveness varies across the operating envelope.
One actuator may be efficient at low speed and weak at high speed. Another may become thermally constrained after sustained use. A control allocator therefore needs current state and may interact with gain scheduling or other mode logic.
An Allocation Diagnostic
- Define the desired system-level control effect.
- List every actuator that can contribute.
- Model each actuator’s effectiveness.
- Include magnitude, rate, thermal and fault constraints.
- Define secondary objectives such as energy, wear or margin.
- Test the allocator near saturation.
- Remove one actuator and test reallocation.
- Verify that priority objectives remain protected.
- Measure whether frequent reallocations create wear, discontinuity or hidden coupling.
The CivDJ Rotation
- Forward: desired control effect → feasible actuator combinations → selected allocation → physical response.
- Backward: start from a failed system-level correction and ask whether available actuator authority was allocated badly or was genuinely insufficient.
- Rotate: inspect the allocation as controller designer, maintenance owner, energy owner, safety owner and receiver.
The lowest-energy allocation may not be the lowest-wear allocation. The lowest-wear allocation may consume reserve margin. The best choice depends on the protected objective and the time horizon.
The Civilisation Lesson
Complex systems rarely have one lever.
The mature task is to separate the required outcome from the particular tool used to produce it, then allocate action across available tools while respecting limits, failures, cost and future margin.
Control allocation is where one required correction becomes a portfolio of feasible actions — not all equal, not all free, and not all available forever.
Return through How Controllability Works, How Control Saturation Works, How Control Systems Work and the master How X Works hub. Together, this Feedback & Control corridor moves from desired state and disturbance to anticipation, tuning, limits, hidden-state knowledge and distributed corrective authority.