⚡ electrical engineering

Monotonic, Minimum-Settling-Time PI Tuning for First-Order-Plus-Dead-Time Plants: A Tangency Characterization

This paper presents an analytical characterization of a PI tuning rule for first-order-plus-dead-time plants that achieves the fastest strictly monotone setpoint response by identifying a non-cancellation operating point defined by a specific tangency identity, offering significant settling time and error improvements over established methods at the cost of increased sensitivity and potential multi-pulse control actions.

Original authors: Senol Gulgonul

Published 2026-06-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Senol Gulgonul

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ 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 driving a car that has a heavy delay between when you press the gas pedal and when the car actually speeds up. This is what engineers call a "First-Order-Plus-Dead-Time" plant. It's like trying to steer a massive ship: you turn the wheel, but the ship doesn't start turning until a few seconds later.

The goal of this paper is to find the perfect way to tune the steering (the PI controller) so that the ship reaches its destination as fast as possible, but with one strict rule: it must never overshoot. If you are parking a robot arm or heating a chemical tank, you can't afford to go past the target and then come back; you must approach it smoothly and stop exactly there.

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

1. The Old Ways vs. The New Way

For a long time, engineers had two main ways to tune these systems:

The "Cancel the Pole" Method: This is like trying to perfectly balance a scale by adding a weight that exactly cancels out the existing weight. It's safe and smooth, but it's slow. It's like driving a car very cautiously to ensure you never overshoot.
The "Triple Pole" Method (MRDP): This tries to make the system as fast as possible by stacking three identical "brakes" together. It works well for some cars, but for ships with long delays, it starts to wobble and overshoot, breaking the "no overshoot" rule.

The Paper's Discovery:
The author found a "sweet spot" that is neither of these. It's a unique tuning that is faster than the safe method and smoother than the fast method.

How it works: Imagine the car has a slow, heavy engine (the real pole) and a fast, bouncy suspension (the complex pair). The old methods tried to kill the bounciness. This new method keeps the bounciness but places a "shock absorber" (the controller zero) right next to the heavy engine. This shock absorber doesn't stop the engine; it just dampens the bounciness just enough so the car never jerks past the target.

2. The "Tangency" Secret

The core of the paper is a mathematical "secret handshake" called the Tangency Identity.

Think of the car's speed as a line on a graph.

If you drive too fast, the line dips below zero (overshoot).
If you drive too slow, the line stays flat.
The "perfect" drive is when the line grazes the zero line like a stone skipping perfectly across a pond. It touches the water but doesn't sink below it.

The paper proves that this perfect "grazing" moment happens at a specific mathematical relationship between the car's speed, the delay, and the steering. This relationship is the "Tangency Identity." It tells us exactly how to set the knobs so the system touches the target and stops, without ever dipping below.

3. The Trade-Off: Speed vs. The "Bump"

The paper is very honest about the cost of this new, faster method. It's not magic; there is a price to pay.

The Good News: The system reaches the target 14% to 52% faster than the old safe methods. It also handles sudden bumps (load disturbances) much better.
The Bad News: To get this speed, the steering wheel (the control signal) has to wiggle a bit.
- For slow, delay-heavy ships: The steering is smooth and perfect.
- For faster ships: The steering does a "two-pulse" move. Imagine you press the gas, then quickly tap the brakes slightly, then press the gas again to settle in. It's a tiny, controlled wiggle.
- The Risk: This wiggle makes the system slightly less robust (more sensitive to noise), but the paper argues this is a fair trade for the massive gain in speed.

4. Why This Matters

The author isn't saying this new method is the "best" for every single situation. Instead, they are saying: "Here is the exact mathematical definition of the fastest possible way to reach a target without overshooting."

If you need the absolute fastest, non-overshooting response, this is the map.
If you need the steering to be perfectly smooth (no wiggles), you stick with the older, slower methods.
If you need the system to be super robust against noise, you stick with the "Cancel the Pole" method.

Summary

The paper finds the "Goldilocks" zone for controlling delayed systems. It uses a clever mathematical trick (the tangency identity) to find a setting where the system is as fast as physically possible without ever going past the target. The only cost is that the control signal might wiggle slightly (a "two-pulse" move) for certain types of systems, but the result is a much quicker and more efficient arrival at the destination.

Technical Summary: Monotonic, Minimum-Settling-Time PI Tuning for FOTD Plants

Problem Formulation
This paper addresses the tuning of a Proportional-Integral (PI) controller for a First-Order-Plus-Dead-Time (FOTD) plant, defined by the transfer function $G(s) = K e^{-Ls} / (Ts + 1)$ . The specific objective is to minimize the 2% settling time ( $T_s$ ) subject to a strict constraint: the plant output must exhibit a strictly monotone (zero-overshoot) step response. Crucially, the control signal itself is left unconstrained regarding its shape (i.e., it is not required to be a "one-pulse" signal). This distinguishes the problem from prior frameworks, such as those by Huba et al., which often couple output monotonicity with a one-pulse control requirement.

The paper positions this optimization against two established reference designs:

Pole-Zero Cancellation (Lambert W): A conservative design that cancels the plant pole, resulting in a clean monotone response but slower settling and poor load disturbance rejection.
Multiple Real Dominant Pole (MRDP): A method placing a triple real pole. While fast for delay-dominated plants, the paper notes that for larger lag ratios ( $T/L$ ), the raw MRDP tuning often violates the monotonicity constraint, leading to overshoot and a two-pulse control signal.

Methodology and Analytical Core
The authors demonstrate that the optimal solution is neither a pole-zero cancellation nor an MRDP design. Instead, the optimizer is a non-cancellation point characterized by a specific modal structure: a slow real pole with a small negative residue, a faster underdamped complex pair, and a controller zero placed near the dominant real pole.

The analytical centerpiece of the work is a tangency identity derived from the condition that the output slope touches zero without crossing (a double zero of the derivative). This identity is expressed as:
$\tan(\omega\tau^\star + \alpha) = \frac{a - b}{\omega}$
where $a$ is the magnitude of the dominant real pole, $-b \pm j\omega$ are the complex poles, and $\tau^\star$ is the time of the "binding dip" where the slope is minimal. This identity establishes that the monotonicity boundary is the locus where the secondary complex mode just fails to drive the output slope below zero, balanced by the slow real mode.

Based on this identity, the two-dimensional search for optimal gains ( $K_p, K_i$ ) collapses into nested scalar conditions, realized at three levels of fidelity:

Explicit Closed-Form Rule: An empirical fit for $K_p$ and $K_i$ as functions of the lag ratio $T/L$ , accurate to within 1.6% of the exact optimum.
Exact Response-Based Reduction: A numerical solver using the true response to find the boundary.
Simulation-Free Transcendental System: A method using a Padé approximation for the delay, reducing the problem to finding the root of a fourth-order polynomial.

Key Results and Performance
The proposed design is compared against the Lambert W cancellation rule and the raw MRDP tuning across a range of lag ratios ( $T/L$ from 0.1 to 10).

Settling Time: The proposed design reduces the 2% settling time by 14% to 52% compared to the cancellation rule. The gains are most significant for delay-dominated plants ( $T/L \lesssim 0.55$ ).
Load Disturbance Rejection: The design lowers the load integrated absolute error ( $IAE_d$ ) by 5% to 38% due to a higher integral gain ( $K_i$ ).
Monotonicity: Unlike the raw MRDP tuning, which exhibits significant overshoot for $T/L > 1.5$ , the proposed design maintains strict output monotonicity across the entire range.
Control Signal Shape:
- For delay-dominated plants ( $T/L \lesssim 0.55$ ), the control signal remains a one-pulse (monotone) signal. In this regime, the design is "Huba-admissible" (satisfying both output monotonicity and one-pulse control) and is strictly faster than the MRDP design.
- For lag-dominated plants ( $T/L \gtrsim 0.55$ ), the control signal becomes two-pulse (exhibiting a slight overshoot and a subsequent dip). While this violates the one-pulse constraint of the Huba framework, the output remains strictly monotone.
Robustness: The speed and load-rejection gains come at the cost of reduced robustness. The maximum sensitivity ( $M_s$ ) increases from the cancellation rule's baseline of 1.39 to a range of 1.44 to 1.62, with the highest penalty occurring in delay-dominated plants.

Significance and Claims
The paper explicitly positions its contribution not as a "uniformly better" tuning rule, but as the exact analytical characterization of a specific, well-defined operating point (the minimum-settling-time solution under output-monotonicity constraints).

The authors claim three primary contributions:

Optimality: They prove that under output-only monotonicity, the standard cancellation and MRDP designs are suboptimal. The true optimum is a non-cancellation design with a specific zero-pole arrangement.
Analytical Characterization: The tangency identity provides an explicit mathematical description of the monotonicity boundary, reducing the search to scalar conditions and offering a closed-form approximation where previous methods relied on numerical performance portraits.
Honest Trade-off Quantification: The paper provides a complete multi-metric comparison, acknowledging that the performance gains are regime-dependent. It clarifies that for delay-dominated plants, the design improves upon MRDP within the admissible set, while for lag-dominated plants, it trades the one-pulse control constraint for faster settling and better load rejection, accepting a modest increase in sensitivity.

The work concludes that the proposed design, the MRDP design, and the cancellation rule are all Pareto-optimal for their respective constraint sets, with this paper supplying the missing analytical point on the performance frontier for the FOTD plant.