Source-linked AI summary
Hybrid Continuous DoA Estimation with Shared-Radius Co-Prime Circular Arrays
Keyvan Aghababaiyan
TL;DR
For continuous 2D-DoA estimation in 3D space, rectangular co-prime arrays [8] are limited by planar geometry and phase ambiguities, while discrete methods suffer grid mismatch when true DoAs miss predefined grid points. The paper combines a shared-radius co-prime circular array with coarse physical-domain search and particle-swarm continuous refinement, jointly processing the sub-array spectra. At -10 dB, the 11-element CCA has approximately 1.05° azimuth RMSE versus 1.9° for the 12-element CRA, and the CCA outperforms it across the tested SNR range.
Problem
For continuous 2D-DoA estimation in 3D space, rectangular co-prime arrays [8] are limited by planar geometry and phase ambiguities, while discrete methods suffer grid mismatch when true DoAs miss predefined grid points.
Source: THIS IS AN AUTHOR-CREATED POSTPRINT… · p. 2, §III · p. 3Method
The paper combines a shared-radius co-prime circular array with coarse physical-domain search and particle-swarm continuous refinement, jointly processing the sub-array spectra.
Source: THIS IS AN AUTHOR-CREATED POSTPRINT… · p. 2, §IV · p. 4Results
At -10 dB, the 11-element CCA has approximately 1.05° azimuth RMSE versus 1.9° for the 12-element CRA, and the CCA outperforms it across the tested SNR range.
Source: §V · p. 8Takeaways & Limitations
Joint processing of both sub-array spectra makes the actual source direction the strictly unique common peak despite ambiguous peaks in individual spectra.
Source: §IV · p. 4Practical hardware implementations still require full-wave electromagnetic calibration for complete 3D antenna radiation patterns and platform scattering.
Source: §2 · p. 3
Abstract
from arXiv · showhide
This paper proposes a shared-radius co-prime circular array for high-resolution, continuous 2D Direction-of-Arrival (DoA) estimation in 3D space, jointly estimating azimuth and elevation angles. The proposed architecture consists of two uniform circular sub-arrays with co-prime antenna counts sharing a common radius RR, a design that intrinsically suppresses mutual coupling leakage compared to dense uniform arrays. Unlike existing works that rely on complex phase-mode transformations to map circular structures to virtual linear arrays, we introduce a hybrid continuous-recovery framework operating directly in the physical spatial domain. By integrating a fast, discrete coarse-grid search with a swarm-intelligence continuous refinement stage, the proposed method completely bypasses discrete grid-mismatch limitations and computationally expensive eigenvalue decompositions. A rigorous theoretical analysis using Nivens Theorem establishes the spatial uniqueness of the true source direction, effectively resolving phase ambiguities. Simulation results demonstrate that this hybrid scheme achieves superior resolution and lower Root Mean Square Error (RMSE) at low Signal-to-Noise Ratios (SNR) compared to uniform configurations, while asymptotically converging to the theoretical Cramer-Rao Bound (CRB) at high SNRs.
I. Introduction
The introduction motivates non-uniform arrays as a route to improved DoA resolution while noting mutual-coupling concerns and gaps in 3D estimation. It presents a shared-radius co-prime circular array and a hybrid continuous-recovery approach as its proposed response.
I. Introduction
- Non-uniform arrays can improve DoA resolution, but closely spaced elements in nested arrays can make them susceptible to mutual coupling.Co-prime linear arrays were introduced as an alternative that reduces coupling while improving estimation accuracy.Source: §I · p. 1
- Linear co-prime methods - are limited to 2D, while rectangular co-prime arrays for 3D estimation can have phase ambiguity.The paper frames joint azimuth-and-elevation estimation in 3D as requiring planar arrays.Source: THIS IS AN AUTHOR-CREATED POSTPRINT… · p. 1
THIS IS AN AUTHOR-CREATED POSTPRINT VERSION
- The proposed shared-radius co-prime circular array targets continuous 2D-DoA estimation in 3D, using physical-domain coarse search and swarm-based continuous refinement.The approach avoids phase-mode transformations and EVD, and is presented as bypassing discrete grid mismatch.Source: THIS IS AN AUTHOR-CREATED POSTPRINT… · p. 2
II. SYSTEM MODEL
The system model uses two co-prime uniform circular sub-arrays with a shared radius and a common reference element. It models incoming narrow-band sources through the array steering response, with mutual coupling represented separately.
II. SYSTEM MODEL
- The array comprises two uniform circular sub-arrays with co-prime element counts and a common radius; they overlap only at the reference element.The total element count is N_T = N_1 + N_2 - 1.Source: §II · p. 2
- The model assumes M uncorrelated narrow-band sources, with azimuth measured from the X-axis and elevation downward from the Z-axis.The received vector is expressed as the steering-matrix response to the source data plus noise.Source: §II · p. 2
- Mutual coupling is modeled by a coupling matrix applied to the array response, whose off-diagonal coefficients depend inversely on element separation.The coupling coefficient also depends on antenna characteristics.Source: §II · p. 2
2. Consequently,
The section argues that the shared-radius co-prime geometry suppresses mutual-coupling leakage relative to a uniform circular array. It also notes that practical implementations still require electromagnetic calibration beyond the geometric model.
2. Consequently,
- The proposed geometry has much lower mutual-coupling leakage than a UCA, so its coupling matrix more closely approximates the identity matrix.The comparison uses the normalized Frobenius norm of the off-diagonal coupling components as the leakage measure.Source: §2 · p. 3
- The geometry bounds and suppresses the summation terms used to characterize coupling leakage.The analysis describes a minimum inter-element distance and says that selecting R appropriately keeps it sufficient to prevent deep-coupling failures seen in dense UCAs.Source: §2 · p. 3
- The distance-dependent geometric model is first-order; practical hardware also needs full-wave electromagnetic calibration for antenna patterns and platform scattering.Source: §2 · p. 3
III. PROPOSED CONTINUOUS DOA ESTIMATION FRAMEWORK
The framework combines a coarse 2D spatial search with continuous swarm refinement to estimate DoA without relying on a fine discrete grid. The coarse stage seeds the refinement, which searches the continuous domain around the peak.
III. PROPOSED CONTINUOUS DOA ESTIMATION FRAMEWORK
- The framework uses a coarse 2D dictionary search followed by continuous swarm refinement to address grid mismatch without EVD or phase-mode transformations.The passage notes that finer grids can mitigate mismatch but exponentially increase computational complexity.Source: §III · p. 3
- The first stage uses correlation matching over a sparse coarse dictionary, with each atom representing a coupled azimuth-elevation pair.This pairing avoids the azimuth-elevation pairing ambiguity described in the passage.Source: §III · p. 3
- The coarse spectrum peak supplies a seed for the particle swarm, whose particles search the continuous cell containing that peak.The algorithm specifies a 2° coarse grid, 40 particles, and 60 maximum iterations.Source: §III · pp. 3–4
IV. DOA ESTIMATION WITH PROPOSED CO-PRIME ARRAY
The method combines spectra from two uniform circular sub-arrays to estimate 2D direction, using their co-prime geometry to make the true direction the unique common peak. A coincidence rule selects that peak, and circular symmetry supports decoupled CRB analysis.
Estimation approach
- The estimator combines spatial spectra from two uniform circular sub-arrays sharing a radius and a physical element at θ=0.The paper motivates this approach as avoiding the high complexity that arbitrary non-uniform array structures can entail.Source: §IV · p. 4
Phase ambiguity
- Larger shared radius produces sharper peaks but introduces phase ambiguity; the analysis restricts elevation to the upper hemisphere and bounds the ambiguity integer.The phase term couples azimuth and elevation in the circular geometry.Source: §IV · p. 4
Uniqueness
- Joint processing makes the true source direction the strictly unique common peak, although each sub-array can display ambiguous grating-lobe peaks.The proof uses the co-prime counts and Niven's Theorem to show nonzero ambiguity integers yield peaks that do not coincide.Source: §IV · p. 4
Coincidence detection
- The coincidence rule selects the closest pair of peak coordinates from the two sub-arrays and discards non-overlapping peaks beyond the grid-resolution tolerance.The final 2D estimate is the pair minimizing Euclidean distance between the peak sets.Source: §IV · pp. 4–5
CRB analysis
- Under the coupling-free CRB model, circular symmetry satisfies the derivative orthogonality conditions, decoupling the bounds; the co-prime design permits a larger radius and lower CRB than standard constrained circular arrays.The bounds are inversely proportional to squared derivative-vector norms.Source: §IV · p. 5
V. EVALUATION
The evaluation reports that the hybrid method improves continuous DoA accuracy while reducing computation, and tests its convergence, snapshot sensitivity, array scaling, coupling leakage, and geometry against alternatives.
V. EVALUATION
- On MATLAB with an Intel Core i5-1335U CPU (1.30 GHz) and 16 GB RAM, the proposed hybrid scheme’s average execution time was approximately 0.15 seconds, versus 0.85 seconds for ESPRIT and 1.45 seconds for standard fine-grid 2D-MUSIC.It operates directly in the physical domain, bypasses EVD, evaluates a reduced coarse grid, and refines the estimate via swarm intelligence.Source: §V · p. 5
- At low SNR, the proposed method has the lowest estimation error; at high SNR, it overcomes MUSIC’s error floor and approaches the theoretical CRB.MUSIC plateaus near 0.4° above 5 dB, while the proposed method reaches near-zero error at 15–25 dB.Source: Figure 3 · p. 6, Figure 4 · p. 6, §V · p. 7
- With 15 snapshots at 0 dB, the proposed method remains robust and significantly outperforms MUSIC, which requires larger samples for reliable covariance estimation.Source: Figure 5 · p. 6, §V · p. 7
- Swarm particles locate the continuous DoA within 10–15 episodes, with steady-state error below 0.05° at 15–25 dB SNR.At −10 dB, errors stabilize at 1.78° for azimuth and 2.05° for elevation; the 2° coarse grid was selected empirically as a trade-off.Source: §V · pp. 5–6, Figure 2 · p. 6
- At 10 dB, a 10-element UCA remains above 0.5° RMSE, while the proposed 6-element co-prime array achieves below 0.2°.The UCA’s error improves as elements are added from 6 to 10, but remains worse than the co-prime array.Source: Figure 6 · p. 7, §V · pp. 7–8
- With 12 elements, coupling leakage energy exceeds 6.2 for the UCA but remains below 3.4 for the co-prime array.The UCA leakage rises from approximately 3.1 at 6 elements as its fixed-radius spacing decreases.Source: Figure 7 · p. 7, §V · p. 8
- Across −10 to 20 dB, the 11-element CCA outperforms the 12-element CRA in azimuth and elevation RMSE.At −10 dB, azimuth RMSE is approximately 1.05° for CCA and 1.9° for CRA.Source: Figure 8 · p. 7, Figure 9 · p. 7, §V · p. 8
VI. CONCLUSION
The paper combines a shared-radius co-prime circular array with hybrid continuous 2D-DoA estimation, reporting spatial uniqueness, reduced mutual coupling, and strong accuracy. Practical deployments still require full-wave electromagnetic calibration for cross-subarray coupling at interleaving points.
VI. CONCLUSION
- The hybrid framework combines a coarse physical-domain search with swarm-intelligence refinement to resolve phase ambiguities and grid mismatch without beam-space transformations or expensive eigenvalue decompositions.Niven’s Theorem confirms spatial uniqueness of the true DoA.Source: §VI · p. 8
- The proposed geometry suppresses mutual coupling leakage, while the method maintains superior low-SNR accuracy and asymptotically converges to the theoretical Cramér-Rao Bound at high SNRs.Source: §VI · p. 8
- Practical deployments must address cross-subarray mutual coupling at interleaving points, typically through full-wave electromagnetic calibration.Source: §VI · p. 8