Source-linked AI summary
Channel Estimation for Extremely Large-Scale Massive MIMO: Far-Field, Near-Field, or Hybrid-Field?
Xiuhong Wei, Linglong Dai
TL;DR
Existing far-field and near-field models do not accurately represent practical XL-MIMO channels containing both path types, limiting direct use of their estimation schemes. The paper proposes a hybrid-field channel model and separately estimates far-field and near-field components; simulations show better NMSE performance with the same low pilot overhead.
Problem
Practical XL-MIMO channels may contain both far-field and near-field paths, whereas existing models assume all scatters belong to one region.
Method
The paper models both path components with an adjustable proportion and estimates them separately using angle-domain and polar-domain transforms.
Results
The proposed scheme achieves better NMSE performance with the same low pilot overhead, while matching near-field and far-field OMP at γ = 0 and γ = 1, respectively.
Takeaways & Limitations
The proposed hybrid-field estimator unifies far-field and near-field channel estimation, with the existing schemes as special cases.
Abstract
from arXiv · showhide
Extremely large-scale massive MIMO (XL-MIMO) is a promising technique for future 6G communications.However, existing far-field or near-field channel model mismatches the hybrid-field channel feature in the practical XL-MIMO system.Thus,existing far-field and near-field channel estimation schemes cannot be directly used to accurately estimate the hybrid-field XL-MIMO channel. To solve this problem, we propose an efficient hybrid-field channel estimation scheme by accurately modeling the XL-MIMO channel.Specifically,we firstly reveal the hybrid-field channel feature of the XL-MIMO channel, where different scatters may be in far-field or near-field region.Then, we propose a hybrid-field channel model to capture this feature, which contains both the far-field and near-field path components. Finally, we propose a hybrid-field channel estimation scheme, where the far-field and near-field path components are respectively estimated. Simulation results show that the proposed scheme performs better than existing schemes.
I. INTRODUCTION
XL-MIMO offers efficiency gains for 6G but creates high-dimensional channel-estimation overhead. Because practical channels can combine far-field and near-field paths, the paper proposes a hybrid-field model and estimation scheme.
- XL-MIMO is promising for 6G because its large antenna array can improve spectral and energy efficiency, but channel estimation incurs unaffordable pilot overhead.
- Far-field schemes model planar-wave channels as sparse in angle and use DFT-based representations with compressive sensing.
- Near-field schemes model spherical-wave channels whose steering vectors depend on both angle and distance, using polar-domain sparsity for low-overhead estimation.
- Practical XL-MIMO channels may contain both far-field and near-field paths, so an all-far-field or all-near-field model mismatches this hybrid-field feature.
- The proposed hybrid-field model includes both path types, controls their proportion with an adjustable parameter, and contains existing models as special cases.
- The proposed estimator separately estimates far-field paths in the angle domain and near-field paths in the polar domain using different transform matrices.
II. SYSTEM MODEL
The system model considers downlink communication from an XL-MIMO base station to a single-antenna user and formulates low-overhead channel estimation from received pilots.
- A. Signal Model: The system uses an N-element XL-MIMO base-station array communicating with a single-antenna user.
- A. Signal Model: The received pilot signal is modeled over M time slots using transmitted pilot signals and additive noise.
- A. Signal Model: Downlink estimation seeks the channel from known received pilots and pilot signals.
- A. Signal Model: Low-overhead estimation targets a pilot count M much smaller than the large antenna count N.
B. Channel Models
The channel environment is divided into far-field and near-field regions according to the distance between the base station and scatterers, with the boundary determined by the Rayleigh distance.
- B. Channel Models: Wireless communication radiation fields are divided into far-field and near-field regions, each associated with a different channel model.
- B. Channel Models: Figure 1 illustrates the near-field and far-field regions used to distinguish the channel-modeling regimes.
1) Far-Field Channel Model:
The far-field channel model applies the planar-wave assumption when scatterers lie beyond the Rayleigh distance, then exploits angle-domain sparsity for low-overhead estimation.
- 1) Far-Field Channel Model:: When a scatterer is farther from the base station than the Rayleigh distance, its far-field channel is modeled using planar waves.
- 1) Far-Field Channel Model:: The far-field model describes each path through its gain and angle, with a planar-wave array steering vector.
- 1) Far-Field Channel Model:: The normalized angle parameter depends on the physical angle and antenna spacing relative to wavelength.
- 1) Far-Field Channel Model:: A DFT matrix transforms the non-sparse spatial channel into an angle-domain representation.
- 1) Far-Field Channel Model:: Limited scatters make the angle-domain channel usually sparse, enabling compressive-sensing estimation with low pilot overhead.
2) Near-Field Channel Model:
Near-field XL-MIMO channels use a spherical-wave model in which array steering depends on both angle and scatter distance. Polar-domain transforms exploit their sparsity, but large transform dimensions and poor column orthogonality can increase energy leakage for far-field components.
- Near-Field Channel Model: Near-field channel modeling uses the spherical wave assumption because steering vectors depend on both angle and scatter distance.The distance is measured from each scatter to the antenna-array center, with antenna-specific path distances determined by the scatter geometry.
- Near-Field Channel Model: The polar-domain transform matrix W consists of near-field steering vectors sampled over joint angle and distance coordinates.Each column corresponds to a sampled angle and distance pair, with multiple sampled distances associated with each sampled angle.
- Near-Field Channel Model: Polar-domain compressed sensing was proposed to exploit near-field sparsity and reduce pilot overhead for near-field channel estimation.The near-field channel is represented as a sparse polar-domain channel under this transform.
- Near-Field Channel Model: The polar transform has a larger dimension and poorer column orthogonality than the DFT matrix, causing more serious energy leakage for far-field channels represented in the polar domain.This contrasts with the DFT matrix, which is associated with angle-domain processing for far-field channels.
- Near-Field Channel Model: Existing far-field and near-field estimation schemes cannot directly and accurately estimate hybrid-field XL-MIMO channels.The limitation arises because practical environments may contain both far-field and near-field components, whereas existing models assume only one region.
III. PROPOSED HYBRID-FIELD CHANNEL ESTIMATION
The proposed estimation framework first characterizes the XL-MIMO channel's hybrid-field nature, then models both field types, and finally estimates them with low pilot overhead.
- III. PROPOSED HYBRID-FIELD CHANNEL ESTIMATION: The proposed scheme reveals the hybrid-field feature, constructs a hybrid-field channel model, and estimates the channel using that model.The stated objective is improved estimation accuracy with low pilot overhead.
A. Hybrid-Field Channel Feature
XL-MIMO environments can contain scatters in both far-field and near-field regions, producing corresponding path components that single-region models cannot capture.
- A. Hybrid-Field Channel Feature: Far-away scatters produce far-field path components, while nearby scatters produce near-field path components.The paper gives a high-altitude base station serving a distant user as an example where the direct link is far-field but nearby scatters generate near-field paths.
- A. Hybrid-Field Channel Feature: Existing far-field or near-field channel models contain only one type of path component and therefore cannot capture the hybrid-field XL-MIMO feature.The hybrid-field environment is defined by the simultaneous presence of far-field and near-field components.
B. Proposed Hybrid-Field Channel Model
The proposed hybrid-field model combines far-field and near-field path components and uses an adjustable parameter to control their proportions. Setting the parameter to either endpoint recovers the standard single-field models.
- B. Proposed Hybrid-Field Channel Model: The hybrid-field channel model includes both far-field and near-field path components with distinct angle-domain and distance-dependent steering vectors.Far-field paths are parameterized by gain and angle, whereas near-field paths additionally include distance.
- B. Proposed Hybrid-Field Channel Model: The adjustable parameter γ controls the proportion of far-field and near-field path components in the model.The model assigns γL far-field components and (1−γ)L near-field components, according to the supplied formulation.
- B. Proposed Hybrid-Field Channel Model: When γ = 1 the model reduces to the standard far-field model, whereas γ = 0 yields a near-field model.Thus, the proposed model treats the existing far-field and near-field models as special cases.
- B. Proposed Hybrid-Field Channel Model: Hybrid-field channels are not sufficiently sparse in either the angle domain or the polar domain because near-field and far-field components spread energy in the mismatched domains.This prevents direct use of existing single-field channel estimation schemes.
C. Proposed Hybrid-Field Channel Estimation
The proposed HF-OMP scheme separately estimates far-field and near-field path components using their respective sparse representations and sensing matrices, then combines them into the channel estimate.
- C. Proposed Hybrid-Field Channel Estimation: HF-OMP formulates separate compressed-sensing problems because hybrid-field channels contain both far-field and near-field path components.The far-field and near-field components are represented sparsely in the angle and polar domains, respectively.
- C. Proposed Hybrid-Field Channel Estimation: The first stage estimates far-field paths in the angle domain with sensing matrix Af = PF, using iterative support selection, least squares, and residual updates.The algorithm performs Lf = γL iterations to identify far-field supports.
- C. Proposed Hybrid-Field Channel Estimation: The second stage estimates near-field paths in the polar domain with sensing matrix An = PW and updates the residual using both estimated path types.Near-field estimation uses Ln = (1 − γ)L iterations, and previously estimated far-field contributions are also removed when applicable.
- C. Proposed Hybrid-Field Channel Estimation: After both stages, the algorithm obtains the full channel by combining the far-field and near-field estimates through their respective transform matrices.The estimated channel is formed from FĥA and WĥP when the corresponding support sets are nonempty.
- C. Proposed Hybrid-Field Channel Estimation: HF-OMP includes existing far-field OMP and near-field OMP as special cases by setting γ = 1 and γ = 0, respectively.The adjustable parameter controls the allocation of estimated components between the two domains.
IV. SIMULATION RESULTS
Simulations evaluate NMSE against SNR and the adjustable parameter γ under specified XL-MIMO settings, including comparisons with far-field OMP, near-field OMP, and MMSE benchmarks.
- IV. SIMULATION RESULTS: N = 512 BS antennas, L = 6 path components, S = 2071 polar-domain grid points, and SNR = 1/σ2 define the simulation setup.The wavelength is λ = 0.01 meters, corresponding to 30 GHz, and path distances are sampled from U(10, 80) meters.
- IV. SIMULATION RESULTS: The proposed HF-OMP scheme is compared with far-field OMP and near-field OMP using M = 256 pilots, while MMSE uses M = 512 pilots and an identity pilot matrix.The transform-grid representation requires estimating 12L non-zero elements in the three sparse-estimation schemes.
- IV. SIMULATION RESULTS: Fig. 4 reports NMSE performance against γ at SNR = 5 dB.The adjustable parameter controls the proportion of far-field and near-field path components in the hybrid-field model.
- IV. SIMULATION RESULTS: At γ = 0, HF-OMP matches near-field OMP NMSE, while at γ = 1, it matches far-field OMP NMSE.These endpoint settings correspond to only near-field or only far-field path components, respectively.
V. CONCLUSIONS
The paper proposes hybrid-field channel estimation for XL-MIMO and reports better NMSE performance than existing schemes with the same low pilot overhead.
- V. CONCLUSIONS: The proposed scheme accurately models the hybrid-field XL-MIMO channel for channel estimation.The conclusion identifies hybrid-field modeling as the paper’s central contribution.
- V. CONCLUSIONS: Existing far-field and near-field channel estimation schemes are special cases of the proposed hybrid-field scheme.The proposed framework therefore encompasses both single-region estimation settings.
- V. CONCLUSIONS: Simulation results show better NMSE performance with the same low pilot overhead.The conclusion reports this outcome without specifying a numerical NMSE value.
- V. CONCLUSIONS: Future work may apply more advanced compressed-sensing algorithms to the hybrid-field channel estimation problem.The paper presents this as a direction for improving performance.