6  Open-Loop and Closed-Loop Systems

6.1 Start With the Missing Measurement

The difference between open-loop and closed-loop control is easiest to see when a measurement disappears. A timed heater can run for ten minutes whether the room is warm or cold; a closed-loop heater checks the temperature and changes course.

IoT teams make this choice constantly when deciding whether a command should trust a schedule, a model, or live feedback. Start by naming the measurement that would prove success. If you cannot measure it in time, the control shape must acknowledge that gap.

In 60 Seconds

Open-loop control sends commands from a schedule, model, or rule without measuring the result for correction. Closed-loop control measures the process variable, compares it with the setpoint, and adjusts the next command from the error. Open loop is useful when the process is predictable and low risk; closed loop is needed when disturbances, drift, or accuracy requirements make correction necessary.

6.2 Learning Objectives

By the end of this chapter, you will be able to:

• Explain the practical difference between command-only control and feedback control.
• Identify when an open-loop design is acceptable for an IoT system.
• Identify when a closed-loop design is required because the system must correct error.
• Describe how disturbances, sensor quality, actuator limits, and placement affect loop design.
• Create a control-loop decision record so later tuning and troubleshooting have useful proof.

Quick Check: Loop Choice Boundaries

MVU: Minimum Viable Understanding

Core concept: Open loop acts from a plan. Closed loop acts from measurement.

Why it matters: A controller cannot correct an error it does not measure. Adding feedback improves adaptation, but it also adds sensor, timing, stability, and fallback responsibilities.

Key takeaway: Many durable IoT systems combine both: an open-loop schedule or model provides a baseline, and a closed-loop correction trims the output when measurements are trustworthy.

6.3 Prerequisites

Before working through this chapter, revisit:

• Feedback Fundamentals: setpoint, process variable, error, and feedback direction.
• Actuators: output devices that turn control decisions into physical action.
• Sensor Fundamentals: measurement quality, sampling, calibration, and sensor failure modes.

6.4 The Two Control Shapes

Open-loop and closed-loop systems differ in one specific design question: does the measured output change the next control command?

Figure 6.1: Open-loop command path

In an open-loop system, the controller applies a command from a schedule, lookup table, operator setting, or model. The output may be inspected by a person or logged by another system, but that measurement is not used by the controller to correct the next action.

Open loop can be a good design when:

• The relationship between command and result is predictable enough.
• A small output error is acceptable.
• The action is reversible or low consequence.
• A feedback sensor would be unreliable, hard to place, or more complex than the value it adds.
• A supervisory system can inspect results at a slower pace.

Figure 6.2: Distributed IoT closed-loop control: a fish-farm dissolved-oxygen example

In a closed-loop system, the measured output is part of the controller input. The controller compares the desired setpoint with the measured process variable, computes an error, and adjusts the actuator command. This is the foundation for on-off, proportional, PI, and PID control.

Closed loop is usually required when:

• The output must stay near a target value.
• Disturbances or loads change during operation.
• The process drifts because of wear, environment, or material variation.
• The system must recover without waiting for a person.
• The team needs proof that the controller reaches and holds the target.

6.5 How to Choose

The choice is not “simple versus smart.” It is a fit decision based on observability, consequence, timing, and expected variation.

Use Open Loop When

• The process is repeatable over the operating range.
• The output does not need tight correction.
• The actuator effect is easy to predict.
• Failure is visible before it becomes severe.
• A slower inspection path is enough.

Use Closed Loop When

• The process variable must be held near a setpoint.
• Disturbances change faster than manual response can handle.
• The actuator response changes with load or wear.
• Overshoot, undershoot, or steady error matters.
• The system must document correction behavior.

Use Hybrid Control When

• A baseline schedule is useful but not sufficient.
• Feedback is available only during some operating states.
• A local loop must keep the process stable while a remote layer adjusts targets.
• A sensor fault should fall back to a known conservative action.
• Commissioning starts with an open-loop test before closing the loop.

6.6 Control Family Addendum

Open loop and closed loop are the first split, but IoT designs often need a more precise control-family label. Use the label only after the measurement, actuator, timing, and fallback proof justify it.

Control family	When it fits
Open-loop schedule	The process is predictable enough, drift is low consequence, and verification can happen at a wider process boundary.
On/off with hysteresis	A binary output can keep the process inside a band, and the hysteresis gap prevents rapid toggling.
Proportional-family control	The actuator accepts variable command strength, and measured error should change command magnitude.
Closed-loop PID candidate	P, I, or D terms each solve an observed loop behavior with trusted measurement, declared sample timing, and output limits.
Hybrid placement	Remote policy or edge coordination sets targets or modes, while local feedback protects fast correction and fallback behavior.

Record the selected family, the controlled variable, actuator authority, expected disturbances, local or remote placement, fallback state, and the trigger for reopening the decision.

6.7 Disturbances and Observability

A disturbance is any influence that changes the process output without being the controller command. Wind, load, supply variation, occupancy, friction, and material changes are common examples. Open-loop control cannot correct a disturbance unless the disturbance is already included in the model or schedule.

Figure 6.3: A closed-loop PID handling disturbances: a clean-room temperature loop holds a setpoint within plus or minus 0.1 degrees Celsius while door openings and equipment heat inject disturbances, and the PID controller senses the changed temperature and adjusts the HVAC command to compensate.

Closed loop does not magically solve every disturbance. It helps only when the loop can observe the output, compute error, and act quickly enough for the process. A feedback design should answer four questions:

• What is measured? Name the process variable, sensor location, sample rate, and quality checks.
• What is corrected? Name the actuator command and the allowed operating limits.
• What changes the process? List expected disturbances and the speed at which they appear.
• What happens on bad data? Define sensor-fault behavior, stale-measurement handling, and fallback state.

6.8 Hybrid Feedback Timing Addendum

Hybrid IoT control is common: a remote service sets policy, an edge layer coordinates nearby equipment, and a local loop performs time-sensitive correction. The design is closed-loop at the physical process only when measured output changes the next local actuation.

Loop item	Proof to preserve
Remote policy	Approved setpoints, schedules, alerts, reports, and fleet comparison, with proof that fast correction does not depend on the remote layer.
Edge coordination	Shared limits, conflict avoidance, gateway or controller ownership, and behavior during wide-area connectivity loss.
Local feedback	Controlled variable, sensor location, sample period, actuator authority, safety limit, and deterministic correction near the process.
Feedback timing	Sample delay, decision delay, actuator response, process response, verification delay, and whether the loop can tolerate missing or noisy samples.
Fallback state	Hold, safe output, alarm, manual mode, limited schedule, or other bounded action when measurement, actuation, or communication is unreliable.
Failure modes	Open-loop drift, bad feedback, oscillation, saturation, repeated alarms, and the trigger that reopens the control-type decision.

A ventilation system may keep a minimum schedule while local feedback extends or reduces fan operation when measured conditions require it. The decision record should compare runtime, process excursions, sensor reliability, and fallback behavior in the actual installation.

6.9 Placement in IoT Systems

Feedback can be local, distributed, or supervisory. The correct placement depends on how quickly the process needs correction and whether the system can continue when the network is unavailable.

Local Loop

• Sensor, controller, and actuator are near the process.
• Best for fast correction and autonomous operation.
• Requires local limits, validation, and tuning records.

Distributed Loop

• Measurement and actuation are separated across devices.
• Useful when one node observes what another node affects.
• Needs message timing, stale-data handling, and conflict rules.

Supervisory Loop

• A slower layer adjusts setpoints, schedules, or operating modes.
• Useful for coordination and long-horizon optimization.
• Should not be the only loop for fast or high-consequence correction.

The common IoT pattern is local closed-loop control for stability, plus a slower supervisory layer for setpoint changes and inspection. A device can also behave open-loop locally while the larger system forms a slower feedback loop through inspection, command updates, or operator checks. The key is to state the loop boundary clearly.

6.10 Control Decision Record

A control decision should leave a short record. This prevents a later maintainer from guessing why a loop was left open, closed locally, or split across layers.

Figure 6.4: Control-loop decision record

Target and Variable

• Setpoint or acceptable range.
• Measured process variable.
• Sensor location and validation rule.

Command and Limits

• Actuator command and units.
• Minimum, maximum, and rate limits.
• Manual override or lockout behavior.

Loop Proof

• Step response or disturbance test.
• Overshoot, settling, steady error, or retry behavior.
• Fallback test for missing or invalid measurement.

6.11 Worked Example: Pumped Tank Level

Consider a tank where a pump adds liquid and an outlet removes liquid at a variable rate.

Open-loop design: The controller runs the pump for a fixed duration every cycle. This works only if the outlet flow and starting level are predictable. If outlet demand changes, the tank may drift high or low because the command does not depend on the measured level.

Closed-loop design: A level sensor reports the tank level. The controller compares level with the target range and adjusts pump output. The loop can correct outlet variation, but it must handle sensor faults, pump saturation, and delayed level response.

Hybrid design: A schedule provides a baseline pump plan, while the level sensor trims the command. If the sensor fails validation, the controller reverts to a conservative open-loop command and raises a maintenance event.

The hybrid design is often the most practical design answer: it uses feedback when the measurement is trusted, and it still has defined behavior when the measurement is not trusted.

6.12 Common Pitfalls

Closing a Loop Without a Reliable Measurement

A noisy, delayed, or misplaced sensor can make a closed-loop system worse than a simple command plan. Validate the measurement before tuning the controller.

Treating Open Loop as Always Wrong

Open loop is appropriate for predictable, low-consequence tasks. The design issue is not whether feedback exists; it is whether missing feedback creates unacceptable error.

Ignoring Actuator Limits

A controller cannot correct an error if the actuator is already at its limit. Record saturation behavior and recovery behavior.

Putting Fast Correction in a Slow Layer

If the process changes faster than messages can travel, keep the stabilizing loop local and use remote layers for slower coordination.

6.13 Knowledge Check

Choose the Loop Type

Match the Control Terms

Order the Loop Decision

Key Concepts

• Open-loop control: Control that applies a command without measuring the output for correction.
• Closed-loop control: Control that uses measured output to reduce error between a target and a process variable.
• Setpoint: The desired target value or target range.
• Process variable: The measured output that represents process behavior.
• Error: The difference between the setpoint and the measured process variable.
• Disturbance: A factor outside the controller command that changes process output.
• Loop placement: The architectural boundary that determines where sensing, control, actuation, and fallback behavior run.

6.14 Overview: Loop Choice Is A Risk Boundary

Open loop, closed loop, and hybrid control are not maturity levels. They are different promises about what the system will notice and correct. An open-loop design promises that the command plan is good enough until inspection. A closed-loop design promises that a measured process variable can safely change the next actuator command.

The risk boundary is the moment the process drifts. If the drift is slow, visible, and low consequence, open loop may be acceptable. If the drift can damage equipment, waste energy, spoil material, or hide until a user is affected, feedback or a bounded hybrid design needs to be part of the architecture.

Use the choice map as an architecture gate: first decide whether output error is acceptable, then decide whether the right variable can be measured quickly enough for correction. Partial measurement, unreliable timing, or fallback requirements point to a hybrid record instead of a vague closed-loop label.

Small-screen read: acceptable drift can stay open loop; unacceptable drift needs measurement; partial or unreliable measurement needs hybrid baseline, feedback trim, and fallback proof.

6.15 Hybrid Loop Record

Many IoT systems should not be labeled simply open loop or closed loop. They combine a baseline schedule, local correction, remote setpoint updates, alarms, and fallback states. Write the record so each layer has a clear job instead of an implied promise.

Baseline Plan

Record the schedule, model, or default command, plus the inspection interval that proves the plan is still acceptable.

Local Correction

Record the measured variable, sensor validation rule, actuator authority, output limit, and stale-measurement behavior.

Supervisory Role

Record which remote layer may change setpoints, modes, alarms, or reports, and which fast corrections must remain local.

A useful record names the rejected alternatives as well. For example, “remote-only feedback rejected because the cabinet temperature can rise faster than the cloud round trip and reconnect path” is more useful than “closed loop selected.”

6.16 Under the Hood: Delay Can Open The Loop

A feedback path is closed only while the measurement, decision, command, and process response remain timely enough. Delay does not merely make a loop slower. It can make the controller act on old state, send repeated corrections after the process has changed, or treat a stale measurement as current proof.

That is why remote supervision is usually safer as a target-setting or reporting layer than as the only stabilizing loop. The local controller can keep a bounded state near the process, while the slower layer changes policy, records evidence, and reopens the decision when operating conditions move outside the original proof.

6.17 Summary

Open-loop systems are simple command paths. They are valid when the process is predictable, consequences are low, and slower inspection is enough. Closed-loop systems measure output and correct error, which makes them better for drift, disturbances, and accuracy requirements. Robust IoT designs often use a hybrid pattern: a baseline command plan, local feedback when measurements are valid, and documented fallback behavior when measurement or communication fails.

6.18 See Also

• Feedback Fundamentals
• PID Control Theory
• Process Control and PID
• Actuators

6.19 What’s Next

Continue with PID System Fundamentals to connect loop architecture with the system-level signals, timing, and loop decisions needed before tuning.