Skip to content
Tutorials Center
My learning

Topics

Lesson 2 of 37 ยท Getting Started

Open-Loop and Closed-Loop Control

Compare open-loop and closed-loop control, see why feedback handles model errors and disturbances, and simulate both in Python with real numbers.

There are two basic ways to control a plant. Open-loop control decides the input from the reference alone and never looks at the result. Closed-loop control, also called feedback control, measures the output and uses the difference from the reference to adjust the input. Almost every serious control system is closed-loop, and this lesson shows exactly why.

Open-loop control

A toaster is a classic open-loop system. You set a timer, the heater runs for that long, and nothing checks whether the bread is actually toasted. A washing machine cycle works the same way.

flowchart LR
    R["reference r"] --> C["controller"]
    C -- "u" --> P["plant"]
    P --> Y["output y"]
    D["disturbance d"] --> P

In an open-loop system the signal flows one way: nothing returns from the output.

Open-loop control is cheap and simple, and it cannot become unstable because of feedback. It works well when the plant model is accurate, nothing disturbs the plant, and the plant itself is stable. Its weaknesses are the same three assumptions failing.

Closed-loop control

Add a sensor and subtract the measurement from the reference. The difference is the error e = r - y, and the controller acts on e.

flowchart LR
    R["reference r"] --> S(("+ -"))
    S -- "error e" --> C["controller"]
    C -- "u" --> P["plant"]
    P --> Y["output y"]
    Y --> M["sensor"]
    M --> S
    D["disturbance d"] --> P

The sensor feeds the output back to the summing junction. This is negative feedback.

The minus sign matters. With negative feedback, a positive error makes the controller push the output up and a negative error pushes it down, so the loop works to remove the error. With positive feedback the effect grows (a microphone squeal), which is useful for oscillators and switches but not for regulation.

Open versus closed on the same plant

Take the first-order plant from the previous lesson with K = 2 and time constant 5 s. We design for the nominal model: to get output 1 open-loop we apply u = 1/K = 0.5. But the real plant has gain 1.6, because heaters age.

Python Model error: open loop versus feedback

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

With the nominal gain the open-loop output is exactly 1. With the aged gain it drops to 0.8, a 20% error, because the controller never notices. The feedback controller ends near 0.95 in both cases. It is not exactly 1, a point we address below, but it is far less sensitive to the plant change.

The reason is simple algebra. At steady state, proportional control with gain kp on a plant with gain K gives y = kp K (r - y), so

Text
y = kp K / (1 + kp K) * r

For kp K = 20 this is 0.952, and for kp K = 16 it is 0.941. A 20% change in K changes the result by only 1%.

Disturbance rejection

Feedback also fights disturbances. Add a constant disturbance d = -0.5 (an open window removing heat) and compare.

Python Rejecting a disturbance

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The disturbance moves the open-loop output from 1.0 to 0.5. The closed-loop output moves only from 0.952 to 0.929, because a falling y increases the error, which increases u.

The costs of feedback

Feedback is not free.

  • Residual error. With only proportional action there is a small steady-state error, which steady-state error analysis quantifies and integral action removes.
  • Stability risk. A loop that corrects too aggressively, especially with delays, can overshoot, oscillate and then grow without bound. Chapter 4 is about preventing this.
  • Sensor noise. The controller reacts to the measurement, noise included.
  • Cost. You need sensors and computation.

Open loop or closed loop?

Open loop Simple, cheap, blind to results.
Closed loop Measures and corrects itself.

Pick open loop only when the plant is predictable.

1 / 2

Feedforward: the best of both

Real designs often combine them. Feedforward control uses a model to compute the input that should be needed (like the r/K term) and lets feedback clean up whatever the model got wrong. A cruise controller can add throttle when it sees a hill coming on the map, and feedback fine-tunes it. Chapter 6 returns to this idea.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which property defines a closed-loop system?

    Closed-loop means the measured output returns to the summing junction so the controller can act on the error. Feedback can in fact cause instability if designed badly.
  2. 2. An open-loop heater is designed for gain 2 but the real gain is 1.6. For a reference of 1, what is the steady output?

    The input is fixed at r/2 = 0.5, so the output is 1.6 x 0.5 = 0.8. Nothing corrects the mismatch.
  3. 3. In negative feedback, the error is defined as

    The measured output is subtracted from the reference, so a positive error pushes the output up.
  4. 4. Which is a genuine drawback of feedback?

    Aggressive feedback, especially with delays, can make the loop unstable. The other options state the opposite of feedback's benefits.

Summary

  • Open-loop control ignores the output; closed-loop control feeds it back and acts on the error e = r - y.
  • Feedback makes the output much less sensitive to model errors and disturbances.
  • Proportional feedback alone leaves a small steady-state error, y = kp K / (1 + kp K) r for a plant with gain K.
  • Feedback can destabilise a system, so stability analysis is essential.
  • Real designs often combine feedforward and feedback.

Next: Block Diagrams and Feedback Basics