⚡ electrical engineering

Parameter Space Design of Repetitive Controllers for Satisfying a Robust Performance Requirement

This paper presents a parameter space design procedure for creating low-order robust repetitive controllers capable of handling plants with time delays, imaginary axis poles, and discontinuous weights, validated through a high-speed atomic force microscope position control example.

Original authors: Burak Demirel, Levent Guvenc

Published 2026-06-04

📖 5 min read🧠 Deep dive

CC BY 3.0

Original authors: Burak Demirel, Levent Guvenc

Original paper licensed under CC BY 3.0 (http://creativecommons.org/licenses/by/3.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to keep a very delicate, wobbly robot arm perfectly still while it is being shaken by a rhythmic, repeating vibration—like a child jumping on a trampoline next to it. The goal is to make the robot arm ignore that shaking and stay exactly where you want it.

This paper presents a new way to design the "brain" (the controller) for such a robot arm. Instead of using complex, heavy mathematical formulas that are hard to visualize, the authors propose a method that maps out a safe zone on a graph.

Here is a breakdown of the paper's ideas using simple analogies:

1. The Problem: The Rhythmic Shaker

In many machines, like high-speed microscopes or hard drives, there are signals that repeat over and over (like a heartbeat). Standard controllers often struggle to cancel out these specific, repeating rhythms perfectly.

The Analogy: Imagine trying to balance a broom on your hand while someone is rhythmically tapping your shoulder. You need a special kind of balance that anticipates that specific rhythm.
The Tool: The paper uses a "Repetitive Controller." Think of this as a special memory loop that learns the rhythm of the disturbance and creates a counter-force to cancel it out perfectly.

2. The Old Way vs. The New Way

The Old Way ( $H_\infty$ methods): The authors compare their method to existing techniques (like $H_\infty$ control) to a black box. You put inputs in, and a complex computer spits out a solution. The resulting controller is often a "monster"—a very complicated, high-order machine that is hard to build and hard to understand. It's like trying to fix a watch with a sledgehammer; it might work, but it's messy and fragile.
The New Way (Parameter Space Design): The authors propose drawing a map. Instead of finding one single "perfect" answer, they find a whole region of safe answers.
- The Analogy: Imagine you are trying to park a car. The old method gives you one specific coordinate (x=5, y=3) and says, "Park exactly here." If you move one inch, you crash.
- The new method draws a green parking lot on the ground. As long as you park your car anywhere inside that green lot, you are safe. This gives the engineer flexibility. If a part of the machine changes slightly, you don't need to redesign the whole thing; you just move your "parking spot" slightly within the safe zone.

3. How the Method Works (The Map)

The authors take the rules for how well the machine must perform (it must be accurate, it must be stable, and it must handle errors) and translate them into a 2D graph.

The Axes: The graph has two axes, representing two adjustable knobs (parameters) on the controller.
The Process: They calculate, for every possible setting of those two knobs, whether the machine would pass the test.
- If a setting passes, it gets colored green (safe).
- If it fails, it stays white (unsafe).
The Result: You end up with a shape (like a blob or an ellipse) on the graph. Any point inside that shape is a valid controller design.

4. Why This is Special

The paper highlights several "superpowers" of this method:

Handling the "Impossible": It can handle machines that have time delays (like a lag in the signal) or unstable parts (poles on the imaginary axis) without breaking the math. It's like being able to drive a car with a broken steering wheel by using a special map that accounts for the wobble.
No "Smooth" Rules Required: Usually, engineers have to pretend that the world changes smoothly and continuously. This method doesn't care. It can handle "jagged" or "discontinuous" rules, which is great for real-world machines that behave unpredictably.
Low Complexity: The resulting controllers are simple and low-order. They aren't giant supercomputers; they are simple, efficient filters that are easy to build.

5. The Real-World Test: The Atomic Microscope

To prove it works, the authors tested this on a High-Speed Atomic Force Microscope (AFM).

The Challenge: These microscopes use a tiny needle to scan surfaces. They need to move incredibly fast and precisely, but the needle vibrates and has its own natural "wobbles" (resonances).
The Setup: They used a specific mathematical model of this microscope. They set up a "safe zone" on their graph that accounted for:
1. Low frequencies: Making sure the microscope tracks the main movement accurately.
2. Middle frequencies: Ensuring it handles the vibrations and errors robustly.
3. High frequencies: Making sure it doesn't become unstable or shake apart at high speeds.
The Outcome: They picked a random point inside the green "safe zone" on their graph. When they simulated the microscope moving in a triangle pattern, it worked perfectly, tracking the path with very little error.

Summary

The paper says: "Stop trying to find the one perfect, complex solution. Instead, draw a map of all the 'good enough' solutions. This gives you a flexible, simple, and robust way to control machines that repeat actions, even if those machines are tricky, delayed, or unstable."

The authors explicitly state this is for Single-Input Single-Output (SISO) systems (one control knob, one sensor) and was demonstrated on an Atomic Force Microscope. They do not claim it works for medical treatments or other unmentioned applications.

Technical Summary: Parameter Space Design of Repetitive Controllers for Satisfying a Robust Performance Requirement

Problem Statement
Repetitive control systems are designed to accurately track periodic reference signals or reject periodic disturbances by utilizing a positive feedback loop containing a time-delay element. While effective, designing these controllers to satisfy robust performance requirements presents significant challenges. Standard robust control methods, such as $H_\infty$ synthesis, struggle with repetitive control systems because the inherent time delay renders the system infinite-dimensional. Consequently, $H_\infty$ approaches often require very high-order weighting functions, leading to high-order repetitive controller filters that necessitate model order reduction for implementation. Furthermore, existing methods often struggle with plants containing time delays, poles on the imaginary axis, or discontinuous weighting functions.

Methodology
The paper proposes a parameter space design procedure that maps robust performance frequency domain specifications directly into the parameter space of chosen repetitive controller filters. The repetitive controller structure consists of a low-pass filter, $q(s)$ , and a dynamic compensator, $b(s)$ , which include time advances to cancel phase lags.

The core of the methodology involves the following steps:

Formulation: The robust performance requirement is defined as $|W_S S| + |W_T T| < 1$ , where $W_S$ and $W_T$ are sensitivity and complementary sensitivity weights, and $S$ and $T$ are the sensitivity and complementary sensitivity functions.
Point Condition Mapping: Using the cosine rule on a geometric representation of the loop gain $L(j\omega)$ , the authors derive a quadratic equation to solve for the magnitude $|L(j\omega)|$ that satisfies the robust performance condition at a specific frequency.
Parameter Extraction: For a chosen frequency, the required loop gain is calculated. This is then mapped back to the parameters of the controller filters $q_p(s)$ and $b_p(s)$ . The filters are structured as second-order rational functions (or products thereof), allowing for the tuning of specific coefficients (e.g., $q_{0i}$ through $q_{5i}$ ).
Region Generation: By iterating through a discrete set of frequencies (fundamental frequency and harmonics), closed curves representing solution boundaries are plotted in the two-dimensional parameter space (e.g., $q_{01}$ vs. $q_{11}$ ).
Intersection: The overall solution region is determined by the intersection of the regions satisfying nominal performance (low frequencies), robust performance (intermediate frequencies), and robust stability (high frequencies).

Key Contributions
The paper distinguishes its approach from existing $H_\infty$ and other robust control methods through several specific advantages:

Visualization and Multi-objective Optimization: The graphical representation in parameter space allows for the easy visualization of solution regions. Multi-objective design is achieved simply by intersecting these regions.
Robustness to Plant Characteristics: The method naturally handles plants with time delays and poles on the imaginary axis without requiring complex infinite-dimensional synthesis techniques.
Discontinuous Weights: Unlike methods requiring continuous frequency weights, this approach accommodates discontinuous weights, as computations are performed only at specific frequencies of interest.
Low-Order Implementation: The method specifies fixed-structure, low-order filters from the outset, avoiding the need for high-order synthesis and subsequent reduction.
Non-Fragile Design: By identifying a solution region rather than a single optimal point, the method facilitates the design of non-fragile controllers where parameter variations within the region do not violate performance objectives.

Limitations
The authors acknowledge specific shortcomings of the proposed method:

It is limited to the simultaneous design of only two controller parameters due to the reliance on graphical 2D visualization.
It does not yield a single analytical solution; instead, it uses a constructive, frequency-by-frequency numerical approach.

Results and Application
The methodology is demonstrated using a high-speed Atomic Force Microscope (AFM) scanner position control example. The plant model includes resonant frequencies and is treated as a fourth-order system.

Design: The authors designed a low-pass filter $q(s)$ and a pure time-advance compensator $b(s)$ . They defined specific sensitivity bounds for low frequencies (nominal performance), intermediate frequencies (robust performance), and high frequencies (robust stability).
Solution Region: A solution region was identified in the $q_{01} - q_{11}$ parameter space by intersecting the constraints.
Simulation: A specific point within the solution region was selected. Simulation results for a triangular wave input at 2 kHz showed that the designed repetitive controller successfully tracked the periodic signal with significantly reduced steady-state error.

Significance
The paper claims that the proposed parameter space method offers a practical alternative to $H_\infty$ methods for repetitive control design. Its primary significance lies in its ability to handle infinite-dimensional systems (due to time delay) and discontinuous weights while producing low-order, easily implementable controllers. The method is presented as a generalization of previous work that focused on nominal performance and robust stability separately; this approach unifies them under a direct robust performance constraint. The study concludes that the method is well-suited for SISO repetitive control systems with discrete frequencies of interest and provides a fast, programmable design procedure.