【Control】⚠ 02. A Case Where FSM-Based Control Did Not Work

topics: [“control”, “FSM”, “PID”]

Introduction

We examined a control configuration in which
FSM (Finite State Machine)–based supervisory control
was added on top of a discrete-time PID control system.

To state the conclusion upfront:

Under the tested conditions, no clear benefit from FSM control was observed.

This article documents that result and provides brief analysis.

🔧 What Was Done

The evaluated configuration was as follows:

Continuous-time plant
Discrete-time PID control (with ZOH assumption)
Introduction of sampling-period jitter
Consideration of disturbances occurring between samples (intersample)
Placement of an FSM as a supervisory layer to switch control modes

The FSM monitored the following indicators:

Growth of control error
Oscillatory tendencies
Variations in control input magnitude

🎯 Expected Behavior

Initially, we expected the following:

FSM intervention when sampling conditions deteriorated
Mode switching during performance degradation
Overall improvement compared to PID-only control

📉 Actual Results

The observed results were:

In many cases, the FSM was not triggered at all
Even when the FSM was triggered, no clear performance improvement was observed
In most conditions, behavior was similar to PID-only control

At least under the tested setup:

We could not claim that “adding an FSM improved control performance.”

🤔 Discussion

Intersample Disturbances Are Outside the FSM’s Scope

Disturbances occurring between samples
are phenomena that alter system states in continuous time.

Since an FSM can only make decisions at sampling instants, it cannot:

Directly suppress such disturbances
Respond immediately at the moment they occur

This is not a design mistake, but a structural limitation of FSM-based control.

FSM Silence Is Not Necessarily a Failure

An FSM is not a device that must always react to every situation.

If:

Control assumptions have not yet collapsed
The lower-layer PID can still handle the disturbance

then it is natural for the FSM to remain inactive.

From this perspective, the result can also be interpreted as:

The FSM correctly avoided unnecessary intervention.

🧾 Summary

FSM-based control was tested, but no clear benefit was observed under the given conditions
Intersample disturbances are not problems that FSMs can directly address
FSMs are not devices for improving control performance, but mechanisms for operational judgment
When applied outside their proper scope, FSMs may appear to “not work”

FSMs and supervisory control are not universal solutions.
Identifying their effective scope is itself part of the design process.

📝 Closing Remarks

Although this is a “case where it did not work,”
the result is still valuable for future design decisions.

For those considering similar architectures,
this may serve as a reminder not to overestimate what FSM-based supervision can achieve.