🔢 mathematics

Reachability Analysis for Design Optimization

This paper presents two methods for approximating reachable sets of linear systems with bounded L-infinity controls, deriving exact characterizations for specific planar systems and demonstrating their practical application in optimizing the design of highly-maneuverable aircraft by incorporating physical maneuver constraints.

Original authors: Steven Nguyen, Jorge Cortés, Boris Kramer

Published 2026-03-18

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Steven Nguyen, Jorge Cortés, Boris Kramer

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 designing a new, super-agile fighter jet. Traditionally, engineers design the shape of the plane (the wings, the fuselage) first, and then, much later, they try to figure out if a pilot can actually fly it through the maneuvers they need. It's like building a car without checking if the engine can handle the speed you want, only to realize too late that the brakes are too weak.

This paper proposes a smarter way: checking the "flight limits" while you are still drawing the blueprints.

Here is a simple breakdown of what the authors did, using some everyday analogies.

1. The Core Idea: The "Reachability Map"

Think of a drone or a jet. If you have a limited amount of fuel or a limit on how hard you can push the throttle (the "control input"), there is a specific area of the sky the aircraft can reach in a certain amount of time.

The Problem: Engineers want to know exactly how big that "reachable area" is so they can design a plane that is guaranteed to be agile enough.
The Challenge: Calculating this area is like trying to draw the exact outline of a cloud. It's mathematically messy, especially when the plane has limits on how hard it can push its engines (which the paper calls $L_\infty$ controls—think of it as a strict "hard cap" on throttle).

2. The Two New Tools

The authors developed two mathematical "flashlights" to shine on this messy cloud and see its shape.

Tool A: The "Switching Switch" (For Simple Planes)

For simpler, flat (2D) systems, the authors found a neat trick. They realized that to push a plane to the very edge of its reachability limit, the pilot (or autopilot) has to push the throttle all the way to the max, then maybe flip it to the min, and stop. It's like a light switch: ON, then OFF.

The Analogy: Imagine you are trying to push a heavy box to the edge of a room. You push as hard as you can, then maybe pull back slightly, but you never push "half-hard."
The Result: Because the control is so simple (just max or min, switching once), they can draw the exact boundary of the reachable area using a simple formula. It's like tracing the shadow of the box perfectly.

Tool B: The "Smoothie Approximation" (For Complex Planes)

For more complex, 3D planes, the "switching" rule gets too complicated. So, the authors used a clever workaround.

The Analogy: Imagine you want to know the shape of a square box (the strict limit), but it's hard to calculate. Instead, you look at a circle, then a hexagon, then a 100-sided shape. As you add more sides, the shape looks more and more like a square.
The Math: They used a mathematical concept called $L_p$ -norms. Think of $L_2$ as a circle and $L_\infty$ as a square. By calculating the reachable area for a "round" version (like $L_6$ ), they found it acts as a very tight, accurate guess for the "square" version (the real-world limit).
The Benefit: It's much easier to calculate the "round" version, and it gives them a very good estimate of the real limits without doing the impossible math.

3. Putting It to the Test: Designing the Jet

The authors didn't just stop at the math; they tested it on a real aircraft design problem.

The Goal: They wanted to design a wing that made the plane more maneuverable (able to reach more places in the sky) without making the plane too heavy or expensive.
The Old Way: They used a standard "energy" metric (like measuring how much fuel it takes to move).
The New Way: They used their new "Reachability Map" as a rule. They told the computer: "Design a wing that increases the size of the reachable sky area by 10%."

The Surprise Result:
When they used the new "Reachability Map" rule, they found a wing design that was much more efficient.

To get a 10% boost in agility using the old method, they had to make the wings 33% bigger (adding a lot of weight).
Using the new method, they only needed to make the wings 0.8% bigger to get that same 10% boost in agility.

Why This Matters

This paper is like giving architects a new ruler. Instead of guessing if a building will stand up after it's built, they can now simulate exactly how the wind will hit the building while they are choosing the materials.

By incorporating these "reachability limits" into the design process, engineers can create aircraft that are naturally more capable and safer, ensuring that the final plane can actually do the cool stunts the mission requires, without needing a last-minute redesign.

In short: They figured out how to mathematically map the "limits of a plane's movement" and used that map to build better, more agile aircraft from the very first sketch.

1. Problem Statement

The paper addresses a critical gap in Multidisciplinary Design Optimization (MDO) for aircraft: the lack of integration between advanced control theory and structural design.

The Issue: Current aircraft design processes often treat control capabilities as an afterthought, optimizing structural parameters (like wing shape) without explicitly accounting for maneuverability constraints. This leads to designs that may be structurally optimal but incapable of executing required mission maneuvers.
The Challenge: Traditional reachability analysis (determining the set of states a system can reach) is computationally expensive, especially for nonlinear systems or those with bounded magnitude inputs ( $L_\infty$ constraints). Furthermore, existing metrics often require specific controller designs, limiting their utility as "controller-agnostic" design metrics.
The Goal: To develop computationally tractable, controller-agnostic reachability metrics based on linear system theory that can be directly incorporated into aircraft design optimization loops to enforce maneuverability constraints.

2. Methodology

The authors propose a two-pronged approach to approximate and characterize reachable sets for linear systems $\dot{x} = Ax + Bu$ under bounded inputs.

A. Exact Characterization for Planar Systems ( $L_\infty$ )

For single-input, planar (2D) systems with real, distinct eigenvalues, the authors derive an exact algebraic characterization of the reachable set boundary.

Optimal Switching: They leverage the property that controls driving the system to the "exposed points" (boundary) of the reachable set are saturated (always at max or min magnitude) and switch at most once.
Parametric Representation: Using this insight, they define two parametric functions, $\xi_1(\eta)$ and $\xi_2(\eta)$ , representing trajectories where the control switches from max to min or min to max at time $\eta$ .
Result: The reachable set $R(T)$ is exactly the convex hull of the images of these two parametric functions. This avoids the need for complex set-propagation algorithms for this specific class of systems.

B. Approximation via $L_p$ -Norms for General Systems

For general linear systems (higher dimensions, multiple inputs), exact $L_\infty$ characterization is difficult. The authors propose approximating the $L_\infty$ reachable set using $L_p$ -norm reachable sets as $p \to \infty$ .

Optimal Control Derivation: They derive the form of the optimal control $u^*(t)$ that minimizes the $L_p$ cost to reach a specific terminal state $x_f$ . The control is shown to be:
$u^*(t) = \left| -B^\top e^{-A^\top t}\lambda^*(0) \right|^{\frac{1}{p-1}} \odot \text{sgn}(-B^\top e^{-A^\top t}\lambda^*(0))$
where $\lambda^*(0)$ is the initial costate vector.
Numerical Approximation: Instead of searching the infinite-dimensional space of control inputs, the method reduces the problem to searching over a grid of initial costate vectors $\lambda(0)$ .
Sufficient Condition: They derive a bound (Proposition 2) relating the $L_p$ norm of the control to the $L_q$ norm of the initial costate (where $q = p/(p-1)$ ). This allows for the generation of an inner approximation of the reachable set by constraining the search space of $\lambda(0)$ , significantly reducing computational cost.

3. Key Contributions

Exact Boundary Characterization: An explicit parametric formula for the boundary of $L_\infty$ reachable sets for single-input, planar linear systems with real, distinct eigenvalues.
$L_p$ -Norm Approximation Framework: A novel numerical method to approximate $L_\infty$ reachable sets by utilizing $L_p$ -norm optimal controls and costate sampling, making the problem tractable for general linear systems.
Controller-Agnostic Metrics: The development of reachability-based metrics (e.g., volume of the reachable set, trace of the Gramian) that do not require a specific feedback controller to be designed beforehand.
Design Optimization Integration: A successful demonstration of incorporating these reachability constraints into a multidisciplinary design optimization problem for a highly maneuverable aircraft.

4. Results

The paper validates the methodology through theoretical proofs and numerical examples:

Planar System Example: For a 2D system, the parametric switching control method perfectly captured the boundary of the reachable set, matching results from established tools like CORA (zonotopic approximation) but providing an exact analytical form.
General System Approximation: Using $p=6$ as an approximation for $L_\infty$ , the authors generated reachable sets that served as tight outer approximations of the bounded-magnitude case. The costate sampling method successfully identified reachable states with high fidelity.
Aircraft Design Optimization Case Study:
- Scenario: Optimizing wing span ( $b$ ) and chord ( $c$ ) to maximize maneuverability (reachable set volume) while minimizing wing perimeter.
- $L_2$ Approach (Baseline): Using the trace of the reachability Gramian as a constraint required a 33% increase in wing perimeter to achieve a 10% increase in maneuverability.
- $L_6$ Approach (Proposed): Using the volume of the $L_6$ reachable set as a constraint achieved the same 10% maneuverability increase with only a 0.8% increase in wing perimeter.
- Implication: The $L_p$ -based metric revealed a much more efficient design trade-off, demonstrating that the $L_2$ (ellipsoidal) approximation was overly conservative for this specific maneuverability goal.

5. Significance

Bridging Control and Design: The work provides a rigorous mathematical framework to embed control capabilities directly into the structural design phase, moving beyond the traditional "design then control" paradigm.
Computational Efficiency: By reducing the search space from control functions to costate vectors and providing exact solutions for specific classes, the method makes reachability analysis feasible for iterative optimization loops.
Design Insight: The case study highlights that different reachability metrics yield vastly different design recommendations. The proposed $L_p$ -based approach offers a more accurate representation of physical limitations (bounded inputs) compared to standard energy-based ( $L_2$ ) metrics, potentially leading to lighter, more efficient aircraft designs.
Future Directions: The authors identify extending these methods to nonlinear regimes and incorporating rate-limited actuator constraints as critical next steps for real-world application.