Design, Modeling, and Validation of Curvature-Based Shape-Aware Flexible Phased Arrays

arXiv:2607.08309 · physics.app-ph · Submitted 2026-07-09 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Design, Modeling, and Validation of Curvature-Based Shape-Aware Flexible Phased Arrays".

Mira: Here's the gist of the paper. Knowledge of the shape of deformed, mechanically flexible phased arrays plays an important role in the ability to maintain their beam pattern<ref:2607.08309#pg2> This work presents the design, modeling, and experimental validation of such arrays with integrated curvature-based shape sensing<ref:2607.08309#pg2> The system utilizes a low-cost, eight-channel, 360◦ digitally controlled phase-shifter platform useful for a wide range of smart-antenna experiments<ref:2607.08309#pg2> The authors demonstrate an end-to-end design of a shape-aware, 6-GHz, eight-element array in conformed and free-hanging scenarios, achieving an average reconstruction accuracy of ∼ 6% while recovering beamsteering performance at radii of curvature below 3.6 cm<ref:2607.08309#pg2> Shape Estimation from Curvature Measurements The method for shape estimation from curvature measurements involves using a sequence of interconnected cubic polynomials to approximate the unknown geometry of a 1-D curve<ref:2607.08309#pg2> The process is similar to spline interpolation, except that in this case the positions of the sample points are unknown<ref:2607.08309#pg2> The polynomials are constrained by the measured curvature at each sample point and by the arc lengths between them which are known design parameters<ref:2607.08309#pg2> The system of equations can be solved numerically using an algorithm presented in

32: , which is readily available in standard scientific programming suites<ref:2607.08309#pg2> This formulation accounts for all measurements simultaneously, thereby enhancing robustness under measurement noise compared to previous segmentation-based shape estimation methods<ref:2607.08309#pg2> Data Sampling and Sensor Placement To evaluate the effectiveness of shape reconstruction, the authors define the reconstruction error Errav as the average Euclidean distance between the coordinates of the sampled points and their estimated positions, normalized by the operating wavelength λ<ref:2607.08309#pg2> The analysis reveals that for both 2nd- and 3rd-order Bezier curves, both the average and the MED of Errav decrease monotonically with the sampling density, as expected<ref:2607.08309#pg2> For N=9, the weighted average of the errors in Figs.

Kai: 4a and 4b is Errav ≈ 5% and MED≈ 10%, yielding performance similar to or better than that of other published shapereconstruction methods<ref:2607.08309#pg2> The reconstruction algorithm is expected to maintain reasonable performance under moderate measurement error of σ=0.05, with Errav ≈ 8% and MED≈ 15%<ref:2607.08309#pg2> Physical Modeling of Sensing Mechanism The physical model relating measured strain to local board curvature was developed and experimentally validated<ref:2607.08309#pg2> The relationship derived is expressed as ∆L/L = h · k, where ∆L/L is the ratio of change in length to original length, h is the board thickness, and k is the curvature<ref:2607.08309#pg2> Strain gauges with a strain limit of 2% allow for measuring a minimum curvature radius of 1 cm in 0.2 mm-thick circuit boards without permanent mechanical damage<ref:2607.08309#pg2> The expected differential voltage can be expressed as ∆Vout = VDD2h · k, which relates the output voltage difference to the curvature<ref:2607.08309#pg2> Experimental Validation and Performance The first experiment involved conformed arrays, where the average Errav over all twelve tested curves was 6.1%, with only one reconstruction exhibiting Errav > 10%<ref:2607.08309#pg2> The third column from the left in Fig.

Mira: 17 shows the raw sensor-curvature readings compared to those calculated from the analytical curves<ref:2607.08309#pg2> In the second experiment with air-suspended arrays, the average reconstruction error over the three inspected shapes was 9.5%, slightly worse than that obtained for the solid surfaces in the previous experiment<ref:2607.08309#pg2> The broadside and first sidelobes are similar when the two different datasets were used for the reconstruction for all the shapes under test, implying that sidelobes do not stem from curvature-based reconstruction errors<ref:2607.08309#pg2> Beam Correction Mechanism The phase adjustment required at each element to compensate for positional variation due to bend is calculated using equations derived from path delay<ref:2607.08309#pg2> The phase delay φi that requires compensation and the path delay ∆i are related by φi/2π = ∆i/λ<ref:2607.08309#pg2> This procedure compensates for deformation effects in flexible arrays by adding a phase shift to the radiated signal<ref:2607.08309#pg2> The resulting radiation patterns show that the power in the steered beam remains within 1.5 dB of that at broadside for most shapes after phase correction<ref:2607.08309#pg2> This work extends and demonstrates a method for reconstructing the shape of flexible phased arrays from local curvature measurements using strain gauges<ref:2607.08309#pg2> The method was demonstrated using a custom-designed flexible antenna array with embedded curvature sensors and low-cost phase shifters, enabling endto-end recovery of beam-steering performance under shape deformation for both rigid-surface and naturally occurring free-hanging geometries<ref:2607.08309#pg2> The calculated average reconstruction error of ∼ 6% is comparable to that of previously introduced methods, while enabling operation at significantly smaller bend radii below 3.6 cm<ref:2607.08309#pg2> Since the method does not require inter-element measurements or iterative feedback, it is attractive for standalone use in physically challenging environments, for single-node sensor applications, and as a complement to other recovery methods<ref:2607.08309#pg2> The approach is general sensing modality applicable beyond the beam correction demonstrated here<ref:2607.08309#pg2> ACKNOWLEDGMENTS The authors would like to thank the Israeli Science Foundation (ISF) for supporting this work under grant No.

Kai: 1234/24<ref:2607.

Title and authors: Kai: So we’re talking about "Design, Modeling, and Validation of Curvature-Based Shape-Aware Flexible Phased Arrays" again. We’ve established that this paper focuses on integrating curvature sensing directly into these flexible arrays to maintain their beam pattern.

Mira: Yeah, the authors present the design and modeling work for a six gigahertz eight element array that works in both conformed and free hanging scenarios, focusing on using curvature-based shape sensing.

Lev: The key thing here is that they are not just looking at how to correct a deformed beam; they are building the sensing mechanism itself into the array design.

Kai: Right, and the paper details how this system utilizes a low cost eight channel three hundred sixty degree digitally controlled phase shifter platform for smart antenna experiments.

Mira: And their shape estimation method is based on using a sequence of interconnected cubic polynomials to approximate the unknown geometry of a one dimensional curve, which they do in an interesting way.

Lev: That sounds like it’s similar to spline interpolation, but with the twist that the positions of the sample points are unknown, and those polynomials are constrained by measured curvature and known arc lengths.

Kai: They solve this system numerically using an algorithm from reference thirty-two which is readily available in standard scientific programming suites <ref:2607.08309#pg3,32, which is readily available in standard scientific programming suites>.

Mira: And they claim this formulation accounts for all measurements simultaneously, which makes it more robust under measurement noise than earlier segmentation based shape estimation methods.

Lev: Robustness is good because we always have some noise in the strain gauge data when you try to run this on real hardware, and we want things that handle that well.

Kai: They also define a reconstruction error Errav as the average Euclidean distance between the coordinates of the sampled points and their estimated positions, normalized by the operating wavelength lambda.

Mira: The analysis shows that for both second and third order Bezier curves, both the average and MED of Errav decrease monotonically with sampling density, which is what we expect.

Kai: For N equals nine, they found that the weighted average of the errors in figures 4a and 4b is Errav approximately five percent and MED approximately ten percent.

Mira: That performance is similar to or better than other published shape reconstruction methods, which is what we expect when you increase the sampling density enough.

Lev: That's a concrete number for us; an error of five percent with that level of sensor density sounds like something we could work towards in a practical setting.

Mira: That's the caveat they put on the reconstruction; it tells us how much noise we can expect before things start getting really noisy.

Lev: And that suggests we need to ensure our strain gauges are accurate enough to stay within that sigma level for reliable shape recovery.

Kai: Moving over to the physical modeling, they developed and experimentally validated a physical model relating measured strain to local board curvature, which they express as delta L over L equals h times k.

Lev: They mentioned that strain gauges with a limit of two percent allow for measuring a minimum curvature radius of one centimeter in zero point two millimeter thick circuit boards without permanent mechanical damage.

The paper's summary: Kai: So, to summarize the paper "Design, Modeling, and Validation of Curvature-Based Shape-Aware Flexible Phased Arrays," they present an integrated system for sensing and recovery.

Mira: They show how this system uses integrated curvature based shape sensing to maintain beam pattern integrity in deformed arrays.

Lev: Essentially, they take physical measurements from strain gauges, model that strain to get curvature, and then use that curvature information to reconstruct the array's shape.

Kai: The system utilizes a low cost eight channel three hundred sixty degree digitally controlled phase shifter platform useful for smart antenna experiments.

Mira: And they demonstrate an end-to-end design of a shape aware six gigahertz eight element array in conformed and free hanging scenarios.

Lev: That means you can have the whole system working from sensing to beam correction in one integrated design, which is a real thing we need for hardware development.

Kai: They achieve an average reconstruction accuracy of about six percent while keeping the beam steering performance at radii below three point six centimeters <ref:2607.08309#pg1,an average reconstruction accuracy of>.

Mira: And they show that this works across both conformed and free hanging scenarios, which is important because it shows generality in the application.

Lev: That's a solid result for us; seeing it work in two different physical setups suggests the sensing mechanism isn't overly dependent on one specific geometry.

Kai: They also developed and validated a physical model relating measured strain to local board curvature, which they express as delta L over L equals h times k.

Mira: That model is what connects the measurable voltage change directly to the curvature of the board, which is what we need for sensing.

Lev: So you’ve got a chain going from physical deformation to electrical signal to geometric information.

The paper's improvements: Kai: The authors suggest several improvements to the original work to make it more useful for real hardware.

Mira: They point out that optimizing over a finite set of spatial probes without incorporating information about the array’s physical shape can yield a suboptimal beam pattern and elevated sidelobe levels.

Lev: That suggests that if we need to incorporate shape information, we shouldn't just rely on some fixed sensor placement; the placement needs to be smarter.

Kai: The second approach they propose is estimating the unknown array shape by measuring the mutual coupling between array radiators.

Mira: Leveraging transceivers already integrated into the array, this method requires neither external feedback nor additional hardware, making it suitable for single node applications such as standalone environmental sensors.

Lev: That sounds like a way to get shape information without needing extra components in the measurement path, which is a big win for low overhead.

Kai: They flag that electromagnetic coupling is sensitive to changes in the surrounding medium, which are not easily distinguishable from changes in element positions in the measured response.

Mira: And they state that multipath reflections, potential surfacewaves, and other interference degrade the signal-to-noise ratio SNR in coupling measurements.

Lev: That’s a fair warning; we need to know how much redundancy we need to plan for these environmental effects before we can deploy this in real hardware.

Kai: The reconstruction algorithm is designed to maintain performance under moderate measurement error of sigma equals zero point zero five, with Errav approximately eight percent and MED approximately fifteen percent.

Lev: So for our real hardware development, we need to ensure our sensors stay within that sigma level for reliable shape recovery.

Conclusion: Kai: So wrapping up the discussion on "Design, Modeling, and Validation of Curvature-Based Shape-Aware Flexible Phased Arrays," the paper shows a method for reconstructing array shape from local curvature measurements.

Mira: It demonstrates an end-to-end design for a six gigahertz eight element array that recovers beam steering performance under shape deformation for both rigid surface and naturally occurring free hanging geometries.

Lev: The calculated average reconstruction error of about six percent is comparable to previously introduced methods, while enabling operation at significantly smaller bend radii below three point six centimeters <ref:2607.08309#pg1>.

Kai: This method doesn't require inter element measurements or iterative feedback, which makes it attractive for standalone use in physically challenging environments.

Mira: Since the approach is general sensing modality applicable beyond the beam correction demonstrated here.

Lev: It’s a good way to get shape information without needing extra components or complex iterative loops, so I think it's useful as a complement to other recovery methods.

Kai: This work on "Design, Modeling, and Validation of Curvature-Based Shape-Aware Flexible Phased Arrays" provides a complete framework for sensing and recovering array geometry.

Mira: It connects physical deformation to an electrical signal to geometric information in a very direct way.

Lev: The ability to recover beam steering performance at radii below three point six centimeters is the most important engineering metric here, and it sets a clear benchmark for how much we can push the system <ref:2607.08309#pg1>.

Yair Dashevsky, Alon Elgarat, Matan Gal-Katziri

Ben-Gurion University

physics.app-ph

Submitted: 2026-07-09

Updated: 2026-10-29

Comments: Revised and expanded version with more informative abstract and comparison table. 13 pages, 18 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 72/100

The gist: Knowledge of the shape of deformed, mechanically flexible phased arrays plays an important role in the ability to maintain their beam pattern<ref:2607.08309#pg2> This work presents the design…

Key concepts

Shape Estimation from Curvature Measurements
This method approximates an unknown 1-D curve by using a sequence of interconnected cubic polynomials. These polynomials are constrained by the measured curvature at each sample point and known arc lengths between them. A numerical algorithm solves this system to determine the array's geometry, offering robustness against measurement noise.
Physical Modeling of Sensing Mechanism
The relationship between physical strain and board curvature is modeled as $\Delta L/L = h \cdot k$, where $\Delta L/L$ is the change in length ratio, $h$ is the board thickness, and $k$ is the curvature. Strain gauges measure this physical strain, allowing the system to calculate the local curvature of flexible boards.
Beam Correction Mechanism
To counteract positional variations caused by bending, a phase shift is added to each element's signal. This compensation uses path delay equations ($\phi_i/2\pi = \Delta i/\lambda$) derived from the bend geometry. This phase adjustment restores the intended beam steering performance, keeping power within 1.5 dB of broadside.
Reconstruction Error (Errav)
Errav is defined as the average Euclidean distance between sampled points and their estimated positions, normalized by wavelength. Analyzing this error shows that reconstruction accuracy improves as sampling density increases, reaching about 6% for N=9 measurements, comparable to other shape reconstruction methods.

Terminology

Summary

Knowledge of the shape of deformed, mechanically flexible phased arrays plays an important role in the ability to maintain their beam pattern<ref:2607. p. 008309#pg2> This work presents the design, modeling, and experimental validation of such arrays with integrated curvature-based shape sensing<ref:2607. p. 008309#pg2> The system utilizes a low-cost, eight-channel, 360◦ digitally controlled phase-shifter platform useful for a wide range of smart-antenna experiments<ref:2607. p. 008309#pg2> The authors demonstrate an end-to-end design of a shape-aware, 6-GHz, eight-element array in conformed and free-hanging scenarios, achieving an average reconstruction accuracy of ∼ 6% while recovering beamsteering performance at radii of curvature below 3. p. 06 cm<ref:2607. p. 0 p. 008309#pg2>

Shape Estimation from Curvature Measurements

The method for shape estimation from curvature measurements involves using a sequence of interconnected cubic polynomials to approximate the unknown geometry of a 1-D curve<ref:2607.08309#pg2> The process is similar to spline interpolation, except that in this case the positions of the sample points are unknown<ref:2607. p. 108309#pg2> The polynomials are constrained by the measured curvature at each sample point and by the arc lengths between them which are known design parameters<ref:2607. p. 108309#pg2> The system of equations can be solved numerically using an algorithm presented in [32], which is readily available in standard scientific programming suites<ref:2607. p. 208309#pg2> This formulation accounts for all measurements simultaneously, thereby enhancing robustness under measurement noise compared to previous segmentation-based shape estimation methods<ref:2607. p. 208309#pg2>

Data Sampling and Sensor Placement

To evaluate the effectiveness of shape reconstruction, the authors define the reconstruction error Errav as the average Euclidean distance between the coordinates of the sampled points and their estimated positions, normalized by the operating wavelength λ<ref:2607.08309#pg2> The analysis reveals that for both 2nd- and 3rd-order Bezier curves, both the average and the MED of Errav decrease monotonically with the sampling density, as expected<ref:2607. p. 308309#pg2> For N=9, the weighted average of the errors in Figs. p. 3 4a and 4b is Errav ≈ 5% and MED≈ 10%, yielding performance similar to or better than that of other published shapereconstruction methods<ref:2607. p. 308309#pg2> The reconstruction algorithm is expected to maintain reasonable performance under moderate measurement error of σ=0. p. 305, with Errav ≈ 8% and MED≈ 15%<ref:2607. p. 308309#pg2>

Physical Modeling of Sensing Mechanism

The physical model relating measured strain to local board curvature was developed and experimentally validated<ref:2607.08309#pg2> The relationship derived is expressed as ∆L/L = h · k, where ∆L/L is the ratio of change in length to original length, h is the board thickness, and k is the curvature<ref:2607. p. 408309#pg2> Strain gauges with a strain limit of 2% allow for measuring a minimum curvature radius of 1 cm in 0. p. 42 mm-thick circuit boards without permanent mechanical damage<ref:2607. p. 408309#pg2> The expected differential voltage can be expressed as ∆Vout = VDD2h · k, which relates the output voltage difference to the curvature<ref:2607. p. 508309#pg2>

Experimental Validation and Performance

The first experiment involved conformed arrays, where the average Errav over all twelve tested curves was 6.1%, with only one reconstruction exhibiting Errav > 10%<ref:2607. p. 908309#pg2> The third column from the left in Fig. p. 9 17 shows the raw sensor-curvature readings compared to those calculated from the analytical curves<ref:2607. p. 908309#pg2> In the second experiment with air-suspended arrays, the average reconstruction error over the three inspected shapes was 9. p. 95%, slightly worse than that obtained for the solid surfaces in the previous experiment<ref:2607. p. 908309#pg2> The broadside and first sidelobes are similar when the two different datasets were used for the reconstruction for all the shapes under test, implying that sidelobes do not stem from curvature-based reconstruction errors<ref:2607. p. 908309#pg2>

Beam Correction Mechanism

The phase adjustment required at each element to compensate for positional variation due to bend is calculated using equations derived from path delay<ref:2607.08309#pg2> The phase delay φi that requires compensation and the path delay ∆i are related by φi/2π = ∆i/λ<ref:2607. p. 708309#pg2> This procedure compensates for deformation effects in flexible arrays by adding a phase shift to the radiated signal<ref:2607. p. 708309#pg2> The resulting radiation patterns show that the power in the steered beam remains within 1. p. 95 dB of that at broadside for most shapes after phase correction<ref:2607. p. 908309#pg2>

Conclusion and Applicability

This work extends and demonstrates a method for reconstructing the shape of flexible phased arrays from local curvature measurements using strain gauges<ref:2607.08309#pg2> The method was demonstrated using a custom-designed flexible antenna array with embedded curvature sensors and low-cost phase shifters, enabling endto-end recovery of beam-steering performance under shape deformation for both rigid-surface and naturally occurring free-hanging geometries<ref:2607. p. 908309#pg2> The calculated average reconstruction error of ∼ 6% is comparable to that of previously introduced methods, while enabling operation at significantly smaller bend radii below 3. p. 96 cm<ref:2607.08309#pg2> Since the method does not require inter-element measurements or iterative feedback, it is attractive for standalone use in physically challenging environments, for single-node sensor applications, and as a complement to other recovery methods<ref:2607. p. 108309#pg2> The approach is general sensing modality applicable beyond the beam correction demonstrated here<ref:2607. p. 008309#pg2>

ACKNOWLEDGMENTS

The authors would like to thank the Israeli Science Foundation (ISF) for supporting this work under grant No. 1234/24<ref:2607. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Shape-aware beam steering in dynamic environments: AI systems can actively compensate for mechanical deformation by using a curvature-based sensing modality to recover an array's shape from low-rate surface measurements, achieving beamsteering performance at radii of curvature below 3.6 cm.

  2. General shape reconstruction capability: The system can reconstruct the shape of flexible phased arrays from local strain data without requiring inter-element measurements or iterative feedback, making it applicable to standalone use in physically challenging environments.

  3. Robust sensor placement optimization: AI algorithms can determine the sensor density required for effective deformation sensing under measurement noise by analyzing a large set of representative deformed shapes, ensuring optimal sensor placement for accuracy.

  4. Noise-robust reconstruction: The reconstruction algorithm is designed to maintain performance under moderate measurement error, as demonstrated when a moderate measurement error of sigma=0.05 results in an Errav approximately 8% and MED approximately 15%.

Related papers