Introduction
I’m working as a control engineer, and majored control theory at university.
Since I started to work with current job, sometimes I felt gap between the theory and the practice – how to approach to real systems. So I’d like to show some impressions from my standpoint.
(I also posted some videos on YouTube)
Feedback control is everywhere.
From temperature regulation to motor speed control, it is often the first approach engineers learn — and for good reason.
But in real-world systems, especially fast ones, feedback alone is often not enough.
In fast systems, waiting for error is already too late.
Feedback Control Is Not Just PID
When people talk about feedback control, PID is often the first thing that comes to mind.
However, PID is just one example — not the definition of feedback control.
In practice, feedback can take many forms:
- State feedback
- Observers
- Optimal control
Despite these variations, all feedback-based approaches share a fundamental characteristic:
They react after a deviation has occurred.
This reactive nature becomes a critical limitation in fast systems.
Why Delays Dominate Fast Systems
In many real-world applications, system dynamics operate on very short timescales.
For example, in automotive transmission control:
- Communication and computation delays: on the order of tens of milliseconds
- Actuator response (e.g., hydraulic pressure, clutch torque): tens to hundreds of milliseconds
- Total shift event duration: roughly 1 to 2 seconds
When the total event lasts only 1–2 seconds, delays on the order of tens to hundreds of milliseconds are no longer negligible —
they fundamentally shape the control strategy.
In fact:
Feedback is not just delayed — it operates on a timescale comparable to the event itself.
This makes pure feedback control inherently reactive and often too slow to achieve high-quality performance.
Real-World Example: Transmission Shift Control
Consider a gear shift in an automatic transmission.
During a shift, torque must be carefully transferred between clutches.
If the torque balance is not properly managed, it results in:
- Shift shock
- Torque holes
- Driveability issues
A purely feedback-based approach would attempt to correct these issues after they occur:
- Detect torque deviation
- Adjust actuator commands
- Wait for system response
But by the time the correction takes effect, the undesirable behavior has already been felt.
You don’t fix a bad shift — you prevent it.
This is why relying solely on feedback is not sufficient in such systems.
The Role of Feedforward Control
To handle fast dynamics, engineers rely heavily on feedforward control.
Unlike feedback, feedforward acts before an error appears.
It uses models, predictions, and known system behavior to anticipate what should happen.
In transmission control, this includes:
- Predicting required torque transfer
- Scheduling clutch pressure in advance
- Coordinating engine and transmission torque
In other words:
Feedforward handles the main action, based on prediction rather than reaction.
Why Feedback Is Still Necessary
Despite its limitations, feedback remains essential.
No model is perfect, and real-world systems are subject to:
- Model uncertainty
- External disturbances
- Temperature variations
- Component aging
Feedback plays a critical role in compensating for these effects.
A practical way to view the relationship is:
Feedforward does the heavy lifting, while feedback handles the imperfections.
Putting It Together
In real-world control systems, the most effective architecture is a combination of both:
- Feedforward for primary control action
- Feedback for error correction
Conceptually:
Target → Feedforward → System → Output
↑
Feedback
As systems become faster, the balance shifts further toward feedforward.
The faster the system, the more you have to rely on prediction rather than reaction.
Conclusion
Feedback control remains a fundamental tool in engineering.
However, in fast systems, it is only part of the solution.
Real-world control strategies rely on a combination of:
- Feedforward (prediction)
- Feedback (correction)
- System modeling
Understanding this balance is key to designing high-performance control systems.
What’s Next
I’m currently working on practical content about real-world control systems, focusing on how these ideas are applied in actual engineering projects.
If you’re interested, stay tuned — I’ll be sharing more insights and examples soon.