【Control】🛡️ 04. Beyond Robust Control: Toward Reliability Control

topics: [“control engineering”, “PID”, “robust control”, “reliability”, “AITL”]

Beyond Robust Control — The Design Philosophy of Reliability Control

For decades, control engineering has revolved around a central question:

“Is the system stable under uncertainty?”

H∞ control, μ-analysis, robust PID, gain margin, phase margin—
all of these are rooted in the design philosophy known as Robust Control.

However, when real systems are operated over long periods of time,
engineers inevitably encounter the following realities:

• Degradation progresses gradually
• Failures are often preceded by warning signs
• Model errors evolve over time
• States exist that are stable, yet practically unusable

At this point, the question shifts beyond mere stability.

“How long can this control system be trusted and used?”

This article proposes Reliability Control
as a design philosophy addressing that question.

🔁 Robust Control vs. Reliability Control

If Robust Control is about
“building a system strong enough not to break,”

then Reliability Control is about
“detecting deterioration and keeping the system usable.”

🧩 What Is Reliability Control?

In this work, Reliability Control is defined as follows:

Reliability Control is a control design philosophy that assumes
degradation and variation in the plant, controller, and environment,
and aims to avoid, delay, or gracefully degrade functional loss.

The key points are:

• It is not optimization
• It does not aim to maintain perfect performance
• Its objective is to preserve usability

🧠 The AITL Controller A-Type Approach

To explore this concept concretely,
the author has initiated preliminary studies using
AITL Controller A-Type, a three-layer control architecture:

• PID × FSM × LLM

👉 AITL Controller A-Type (GitHub Pages)
https://samizo-aitl.github.io/aitl-controller-a-type/

The AITL Controller A-Type consists of:

• PID: real-time stabilization (V–I response control)
• FSM: state supervision, mode transitions, degraded operation
• LLM: redesign and decision-making based on degradation or anomaly signs

This structure embeds Robust Control in the inner loop
while constructing an outer Reliability Control layer.

📊 Initial Experiment: Comparison After 1000 Days of Wear

Using a model that assumes wear and degradation,
control responses were compared after 1000 days of equivalent degradation.

Compared Controllers

• Conventional PID control
• AITL control (PID × FSM × LLM)

Evaluation Metric

• Phase deviation relative to the initial (non-degraded) response waveform

The results showed a tendency that,
even after 1000 days of wear, AITL control exhibited a smaller phase deviation
relative to the initial waveform than conventional PID control.

This suggests behavior that attempts not merely to remain stable,
but to preserve response characteristics closer to the original behavior over time.

⚖️ Current Positioning

That said, at the present stage:

• Evaluation metrics are limited
• Degradation models are simplified
• Statistical significance and generality are unverified

Therefore, it is premature to conclude that AITL control is truly effective
as Reliability Control.

At this point, the results merely indicate that:

A behavior different from traditional Robust Control may be emerging.

🔭 Future Work and Outlook

Future investigations will focus on:

• Diversifying wear and degradation models
• Expanding reliability metrics
  (response delay, amplitude degradation, etc.)
• Separating the contributions of FSM transitions and LLM redesign
• Quantitative comparison with “stable but unusable” states

Through these efforts,
the goal is to assess whether AITL control meaningfully embodies
the principles of Reliability Control.

📝 Closing Remarks

Robust Control should not be rejected.
It will continue to serve as the foundation of control systems.

At the same time:

Is control engineering not, ultimately,
a technology for continuing operation in a world that inevitably degrades?

Reliability Control emerges as a design philosophy
that stands beyond the culmination of Robust Control.

If this article serves as an entry point
to that discussion, it has achieved its purpose.