betaFeedback

WorksEngineering

Geometry-Aware Channel Aging in MIMO Systems: When Is Kronecker Separability Enough?

Abstract

This letter investigates when Kronecker-separable spatial correlation approximates the covariance of multiple-input-multiple-output (MIMO) channels. We propose a geometry-aware channel model in which the same propagation environment determines both the Weischelberger spatial coupling matrix and the temporal correlation coefficient. The propagation environment follows a 3GPP-inspired clustered channel model. We consider a MIMO system affected by symbol-level channel aging caused by the user equipment's mobility. We schedule both uplink and downlink pilots, thereby deriving the respective spectral efficiencies (SEs) in the presence of imperfect channel state information. Numerical results show that channel aging's magnitude is strongly geometry-dependent, with diffuse propagation inducing greater temporal decorrelation and SE degradation than concentrated scattering. The results also identify regimes where Kronecker separability can be applied and regimes where geometry-aware Weischelberger coupling is needed.

Introduction

Multiple-input multiple-output (MIMO) systems rely on accurate channel state information (CSI) to perform multi-user (MU)-precoding and interference suppression. In practice, CSI is acquired through pilot transmission and becomes progressively outdated over time due to user equipments (UEs) mobility and the dynamic evolution of the propagation environment. This phenomenon, commonly referred to as channel aging [1], becomes prevalent in high-mobility scenarios, where the channel variations within the frame may become non-negligible [2], resulting in a degradation of the achievable spectral efficiency (SE). The channel’s temporal correlation decay is often modeled using Jake-like isotropic scattering models, which produce correlation coefficients mainly determined by the UE velocity magnitude [3]. Though mathematically tractable, this model fails to capture the effect of the propagation geometry, as UEs with the same velocity magnitude may exhibit different correlation decays depending on the positions of their dominant multipath clusters.

The propagation geometry is represented by the channel spatial covariance matrix. Thanks to its mathematical tractability, the Kronecker-separable model is among the most widely adopted [4]. This model represents the MIMO channel covariance as the product of one-sided base station (BS) and UE correlation matrices. This separability assumption neglects the joint correlation between BS and UE, which is especially important in MU-MIMO networks. The Weischelberger [5] model overcomes this limitation by introducing a coupling matrix that describes how power is distributed across BS-UE eigenmode pairs [6]. Nevertheless, this coupling is often treated as an abstract entity rather than being explicitly linked to the propagation geometry [7].

The need for a unified framework where the same propagation environment gives rise to both the spatial and the temporal correlation of the MIMO channel motivates the main question in this letter: When is the Kronecker-separable model sufficiently accurate, and when is joint-correlation coupling modeling necessary to capture the propagation geometry?

Contributions

This letter proposes a geometry-aware MIMO channel-aging model that links spatial eigenmode coupling and temporal correlation through the same propagation geometry. The propagation geometry is inspired by the 3GPP-standardized cluster delay line channel model. The propagation environment is indeed made of propagation clusters, whose positions are randomly generated, even though it is worth mentioning that the same framework could be driven by ray tracing in practical deployments We construct a Weischelberger coupling matrix by projecting the cluster-level covariance contributions onto the BS and UE eigenmodes. A cluster-dependent temporal correlation coefficient is then derived in which each cluster Doppler shift is weighted by its power, UE-side angle, and Doppler spread. To this end, we propose a Doppler-spread model that directly links the uncertainty in the angular domain to that in the Doppler domain. Finally, we evaluate the symbol-level aging impact in a time-division duplexing (TDD) MIMO system with explicit uplink (UL) and downlink (DL) pilots, covariance-based channel estimation, and imperfect-CSI SE expressions. Numerical results identify regimes where the Kronecker approximation is accurate and regimes where geometry-aware coupling is needed, showing that diffuse propagation induces stronger temporal decorrelation and larger SE losses than concentrated scattering.

System Model

We consider a MU-MIMO system, where a BS, located at the origin and equipped with MBSM_{\mathrm{BS}} antennas, serves KK UEs, equipped with MUEM_{\mathrm{UE}} antennas, located at ek=[ek,x,ek,y]T\boldsymbol{e}_k = [e_{k,x},e_{k,y}]^{\mathsf{T}}. The BS performs orthogonal frequency division multiplexing

(OFDM) transmission around the carrier frequency fˉ\bar{f}, over VV active subcarriers with a subcarrier spacing of Δf\Delta_f. The subcarriers are partitioned into BB blocks of VchoV_{\mathrm{cho}} subcarriers, indexed by the block index nn: the channel is assumed constant across subcarriers within the same block. Here T=1/Δf+TcpT = 1/\Delta_f + T_{\mathrm{cp}} is the symbol duration, with TcpT_{\mathrm{cp}} being the cyclic prefix duration. We define the symbol time index ii, with time unit TT. The frequency-domain UL communication channel between UE kk and BS is denoted by Hk,i,n∈CMBS×MUE\mathbf{H}_{k,i,n} \in\mathbb{C}^{M_{\mathrm{BS}}\times M_{\mathrm{UE}}}. TDD and UL/DL channel reciprocity are assumed, i.e., Hk,i,nDL=Hk,i,nT\mathbf{H}^{\mathrm{DL}}_{k,i,n} = \mathbf{H}^{\mathsf{T}}_{k,i,n}. We account for non-negligible time correlation between symbol times: the correlation decay is modelled as a first-order autoregressive (AR) process with a scaled identity state-transition matrix [1]. The channel’s time correlation is thus defined as E[hk,i,nhk,i′,nH]=Ckζk(∣i−i′∣)\mathbb{E}[\mathbf{h}_{k,i,n}\mathbf{h}_{k,i',n}^{\mathsf{H}}] = \mathbf{C}_{k}\zeta_k(|i-i'|), where ζk(∣i−i′∣)\zeta_k(|i-i'|) captures the correlation decay. The frequency domain UL observation at the BS at time ii on a generic subcarrier vv belonging to block nn is defined as

yi,nUL≜∑k=1KαkHk,i,nWk,i,nxk,i,v+ni,v,(1)\mathbf{y}^{\mathrm{UL}}_{i,n} \triangleq\sum_{k=1}^{K}\alpha_k\mathbf{H}_{k,i,n}\mathbf{W}_{k,i,n}\mathbf{x}_{k,i,v}+\mathbf{n}_{i,v}, \tag*{(1)}

where Wk,i,n∈CMUE×MUE\mathbf{W}_{k,i,n} \in\mathbb{C}^{M_{\mathrm{UE}}\times M_{\mathrm{UE}}} is the precoding matrix adopted by UE kk at time ii for all the subcarriers in block nn. Here, αk\alpha_k accounts for the path loss and large-scale fading of UE kk. Here xk,i,v∈CMUE\mathbf{x}_{k,i,v} \in\mathbb{C}^{M_{\mathrm{UE}}} are the transmitted symbols adopted by UE kk in its UL transmission. We assume that xk,i,v∼CN(0,IMUE)\mathbf{x}_{k,i,v} \sim\mathcal{CN}(0,\mathbf{I}_{M_{\mathrm{UE}}}). The BS receiver noise is ni,v∼CN(0,IMUEσBS2)\mathbf{n}_{i,v} \sim\mathcal{CN}(0,\mathbf{I}_{M_{\mathrm{UE}}}\sigma_{\mathrm{BS}}^2). Conversely, the frequency-domain DL observation at UE kk is defined as

yk,i,nDL≜αkHk,b,nDLFi,nxi,v+ni,v,k,(2)\mathbf{y}^{\mathrm{DL}}_{k,i,n} \triangleq\alpha_k\mathbf{H}^{\mathrm{DL}}_{k,b,n}\mathbf{F}_{i,n}\mathbf{x}_{i,v}+\mathbf{n}_{i,v,k}, \tag*{(2)}

where Fi,n=[F1,i,n,…,FK,i,n]∈CMBS×KMUE\mathbf{F}_{i,n} = [\mathbf{F}_{1,i,n},\ldots,\mathbf{F}_{K,i,n}] \in\mathbb{C}^{M_{\mathrm{BS}}\times KM_{\mathrm{UE}}} is the precoder matrix adopted for every subcarrier in the nn-th block. The transmitted symbols are defined as xi,v=[xi,v,1,…,xi,v,K]T∈CMUEK\mathbf{x}_{i,v} = [\mathbf{x}_{i,v,1},\ldots,\mathbf{x}_{i,v,K}]^{\mathsf{T}} \in\mathbb{C}^{M_{\mathrm{UE}}K}, with xi,v,k∼CN(0,IMUE)\mathbf{x}_{i,v,k} \sim\mathcal{CN}(0,\mathbf{I}_{M_{\mathrm{UE}}}). Lastly, ni,v,k∼CN(0,IMUEσk2)\mathbf{n}_{i,v,k} \sim\mathcal{CN}(0,\mathbf{I}_{M_{\mathrm{UE}}}\sigma_k^2) is UE kk receiver noise.

Weischelberger correlation model

We here take the Weichselberger Rayleigh fading model [5] as a reference and use it to study when the Kronecker-separable approximation is accurate. This model introduces an eigenmode coupling matrix, here denoted by Ωk∈R+MBS×MUE\mathbf{\Omega}_k \in\mathbb{R}_{+}^{M_{\mathrm{BS}}\times M_{\mathrm{UE}}}, specifying how the channel power is distributed across BS–UE eigenmode pairs, while the Kronecker model is obtained as a separable special case, where the BS and UE eigenmodes are perfectly uncoupled. Let us define the one-sided correlation matrices of the UL channel between UE kk and BS as CBS,k=E[HkHkH]∈CMBS×MBS\mathbf{C}_{\mathrm{BS},k} = \mathbb{E}[\mathbf{H}_k\mathbf{H}_k^{\mathsf{H}}] \in\mathbb{C}^{M_{\mathrm{BS}}\times M_{\mathrm{BS}}} and CUE,k=E[HkDLHkDL,H]∈CMUE×MUE\mathbf{C}_{\mathrm{UE},k} = \mathbb{E}[\mathbf{H}^{\mathrm{DL}}_k\mathbf{H}_k^{\mathrm{DL},\mathsf{H}}] \in\mathbb{C}^{M_{\mathrm{UE}}\times M_{\mathrm{UE}}}, with their respective eigendecompositions CBS,k=UBS,kdiag⁡(λBS,k)UBS,kH\mathbf{C}_{\mathrm{BS},k} = \mathbf{U}_{\mathrm{BS},k}\operatorname{diag}(\boldsymbol{\lambda}_{\mathrm{BS},k})\mathbf{U}_{\mathrm{BS},k}^{\mathsf{H}} and CUE,k=UUE,kdiag⁡(λUE,k)UUE,kH\mathbf{C}_{\mathrm{UE},k} = \mathbf{U}_{\mathrm{UE},k}\operatorname{diag}(\boldsymbol{\lambda}_{\mathrm{UE},k})\mathbf{U}_{\mathrm{UE},k}^{\mathsf{H}}. The UL channel covariance CkUL=E[vec⁡(H)vec⁡(H)H]\mathbf{C}^{\mathrm{UL}}_k = \mathbb{E}[\operatorname{vec}(\mathbf{H})\operatorname{vec}(\mathbf{H})^{\mathsf{H}}] is defined as

CkUL=(UUE,k∗⊗UBS,k)DΩ,k(UUE,k∗⊗UBS,k)H(3)\mathbf{C}^{\mathrm{UL}}_k = (\mathbf{U}_{\mathrm{UE},k}^{*}\otimes\mathbf{U}_{\mathrm{BS},k})\mathbf{D}_{\mathbf{\Omega},k}(\mathbf{U}_{\mathrm{UE},k}^{*}\otimes\mathbf{U}_{\mathrm{BS},k})^{\mathsf{H}} \tag*{(3)}

Where DΩ,k=diag⁡(vec⁡(Ωk))\mathbf{D}_{\mathbf{\Omega},k} = \operatorname{diag}(\operatorname{vec}(\mathbf{\Omega}_k)). The DL channel covariance CkDL=E[vec⁡(HDL)vec⁡(HDL)H]\mathbf{C}^{\mathrm{DL}}_k = \mathbb{E}[\operatorname{vec}(\mathbf{H}^{\mathrm{DL}})\operatorname{vec}(\mathbf{H}^{\mathrm{DL}})^{\mathsf{H}}] follows by reciprocity:

CkDL=(UBS,k∗⊗UUE,k)DΩT,k(UBS,k∗⊗UUE,k)H(4)\mathbf{C}^{\mathrm{DL}}_k = (\mathbf{U}_{\mathrm{BS},k}^{*}\otimes\mathbf{U}_{\mathrm{UE},k})\mathbf{D}_{\mathbf{\Omega}^{\mathsf{T}},k}(\mathbf{U}_{\mathrm{BS},k}^{*}\otimes\mathbf{U}_{\mathrm{UE},k})^{\mathsf{H}} \tag*{(4)}

For simplicity, the traces of both matrices are normalized to one. It should be noted that the coupling matrix ΩkKron=1MBSMUEλBS,kλUE,kT\mathbf{\Omega}^{\mathrm{Kron}}_k = \frac{1}{M_{\mathrm{BS}}M_{\mathrm{UE}}}\boldsymbol{\lambda}_{\mathrm{BS},k}\boldsymbol{\lambda}_{\mathrm{UE},k}^{\mathsf{T}} yields a Kronecker-separable covariance model. The covariance mismatch between the Kronecker-separable model and the jointly-correlated one is measured as

dΩ=max⁡k∥Ωk−ΩkKron∥F2∥Ωk∥F2.(5)d_{\mathbf{\Omega}} = \max_k \frac{\left\lVert\mathbf{\Omega}_k-\mathbf{\Omega}^{\mathrm{Kron}}_k\right\rVert_{\mathrm{F}}^2}{\left\lVert\mathbf{\Omega}_k\right\rVert_{\mathrm{F}}^2}. \tag*{(5)}

Propagation geometry

The propagation environment between the BS and the UEs is composed of a set of CC scattering clusters. Cluster cc is centered at pc=[px,c,py,c]T\mathbf{p}_c=[p_{x,c},p_{y,c}]^{\mathsf{T}} and is seen by the BS at the angle ψc\psi_c and at the angle φk,c\varphi_{k,c} by UE kk. The angles of the rays scattered by each cluster follow a normal distribution with side-dependent angle spreads, denoted by σBS,c\sigma_{\mathrm{BS},c} and σUE,k,c\sigma_{\mathrm{UE},k,c}. Assume that the size of each cluster is described by its effective radius rcr_c. The angle spreads of those clusters are then computed as

σBS,c=arctan⁡(rc∥pc∥),σUE,k,c=arctan⁡(rc∥pc−ek∥).(6)\sigma_{\mathrm{BS},c}=\arctan\left(\frac{r_c}{\lVert\mathbf{p}_c\rVert}\right),\qquad \sigma_{\mathrm{UE},k,c}=\arctan\left(\frac{r_c}{\lVert\mathbf{p}_c-\mathbf{e}_k\rVert}\right). \tag*{(6)}

The rationale behind this model is that closer clusters scatter rays within a wider angular interval than farther ones. The marginal covariances CBS,k\mathbf{C}_{\mathrm{BS},k} and CUE,k\mathbf{C}_{\mathrm{UE},k} are computed according to the model in [8]: under the assumption of small angle spreads, we can write CBS=∑c=1CρcBSCBS,c\mathbf{C}_{\mathrm{BS}}=\sum_{c=1}^{C}\rho_c^{\mathrm{BS}}\mathbf{C}_{\mathrm{BS},c}, where

[CBS,c]l,m≈e−jπ(l−m)sin⁡(ψc)e−π22((l−m)cos⁡(ψc)σBS,c)2.(7)[\mathbf{C}_{\mathrm{BS},c}]_{l,m}\approx e^{-\mathsf{j}\pi(l-m)\sin(\psi_c)}e^{-\frac{\pi^2}{2}((l-m)\cos(\psi_c)\sigma_{\mathrm{BS},c})^2}. \tag*{(7)}

The definition of CUE,k\mathbf{C}_{\mathrm{UE},k} follows from the same model and is omitted for brevity. Here ρcBS\rho_c^{\mathrm{BS}} represents the power of cluster cc as seen from the BS; conversely, ρc,kUE\rho_{c,k}^{\mathrm{UE}} represents the cluster power seen by UE kk, while ρc,k\rho_{c,k} is the power of said cluster in the coupling matrix. Each of those powers is generated by a geometry-based adaptation of the standardized power delay profile [9]. The clusters BS side, UE side, and joint delays are defined as τˉcBS=∥pc∥/c0\bar{\tau}^{\mathrm{BS}}_c=\lVert\mathbf{p}_c\rVert/c_0, τˉc,kUE=∥pc−ek∥/c0\bar{\tau}^{\mathrm{UE}}_{c,k}=\lVert\mathbf{p}_c-\mathbf{e}_k\rVert/c_0, and τˉc,k=(∥pc∥+∥pc−ek∥−∥ek∥)/c0\bar{\tau}_{c,k}=(\lVert\mathbf{p}_c\rVert+\lVert\mathbf{p}_c-\mathbf{e}_k\rVert-\lVert\mathbf{e}_k\rVert)/c_0, with c0c_0 denoting the speed of light. Each of those delays is then re-normalized such that the closest cluster has zero delay, i.e., τc,k=τˉc,k−min⁡lτˉl,k\tau_{c,k}=\bar{\tau}_{c,k}-\min_l\bar{\tau}_{l,k}. The cluster powers are then computed as [9 p. 54]

ρc,k=exp⁡{−τc,k(rτ−1)/(rτDS)}10−Zc,k/10,(8)\rho_{c,k}=\exp\left\{-\tau_{c,k}(r_\tau-1)/(r_\tau DS)\right\}10^{-Z_{c,k}/10}, \tag*{(8)}

where DSDS and rτr_\tau are the root mean square (RMS) delay spread and the delay proportionality factor. The per-cluster shadowing term is denoted by Zc,k∼Lognorm⁡(0,σZ2)Z_{c,k}\sim\operatorname{Lognorm}(0,\sigma_Z^2). These powers are then normalized such that ∑cρc,kBS=∑cρc,kUE=∑cρc,k=1\sum_c\rho^{\mathrm{BS}}_{c,k}=\sum_c\rho^{\mathrm{UE}}_{c,k}=\sum_c\rho_{c,k}=1, ∀k\forall k. We propose the following coupling matrix

Proposition 2. Let uBS,l\mathbf{u}_{\mathrm{BS},l} and uUE,k,m\mathbf{u}_{\mathrm{UE},k,m} denote the ll-th largest eigenvector of CBS\mathbf{C}_{\mathrm{BS}} and the mm-th largest eigenvector of CUE,k\mathbf{C}_{\mathrm{UE},k}, then the geometry-based coupling matrix

[Ωk]l,m=∑c=1Cρc,kJνc,lυk,c,m(9)[\mathbf{\Omega}_k]_{l,m}=\sum_{c=1}^{C}\rho_{c,k}^{J}\nu_{c,l}\upsilon_{k,c,m} \tag*{(9)}
νc,l=(uBS,lHCBS,cuBS,l)/MBS(10)\nu_{c,l}=\left(\mathbf{u}_{\mathrm{BS},l}^{\mathrm{H}}\mathbf{C}_{\mathrm{BS},c}\mathbf{u}_{\mathrm{BS},l}\right)/M_{\mathrm{BS}} \tag*{(10)}
υk,c,m=(uUE,k,mHCUE,k,cuUE,k,m)/MUE(11)\upsilon_{k,c,m}=\left(\mathbf{u}_{\mathrm{UE},k,m}^{\mathrm{H}}\mathbf{C}_{\mathrm{UE},k,c}\mathbf{u}_{\mathrm{UE},k,m}\right)/M_{\mathrm{UE}} \tag*{(11)}

This coupling matrix maps the propagation environment to eigenmode coupling: [Ωk]l,m[\mathbf{\Omega}_k]_{l,m} is large when both eigenmodes are excited by one or more clusters with high power. How much an eigenmode is excited by a cluster is measured by projecting the cluster contribution to the marginal covariance (e.g., CBS,k,c\mathbf{C}_{\mathrm{BS},k,c}) onto the corresponding eigenvector (e.g., uBS,k,l\mathbf{u}_{\mathrm{BS},k,l}). A Sinkhorn-like scaling is applied to Ω\mathbf{\Omega} [10] to match the row and column sums to the BS and UE marginal eigenvalues, thus ensuring a fair comparison with Ωkron\mathbf{\Omega}^{\mathrm{kron}}.

Temporal correlation

In our previous works, the temporal correlation coefficient ζ\zeta followed a Jake’s model. While analytically convenient, this model averages different propagation directions and does not account for the fact that the UE’s Doppler shift is not isotropic, meaning that clusters at different positions age at different rates. Let denote UE kk velocity with νk=∥νk∥[cos⁡(θk),sin⁡(θk)]\boldsymbol{\nu}_k=\lVert\boldsymbol{\nu}_k\rVert[\cos(\theta_k),\sin(\theta_k)], with θk\theta_k indicating the velocity’s direction. therefore, the Doppler shift seen by cluster cc is defined as fD,k,c=(fˉ∥νk∥/c0)cos⁡(φk,c−θk)f_{D,k,c}=(\bar{f}\lVert\boldsymbol{\nu}_k\rVert/c_0)\cos(\varphi_{k,c}-\theta_k). Following the model presented in [11 Lemma 1], the correlation coefficient is

ζk(i)=∑c=1Cρk,cUEej2πifD,k,cTe−2(πTiςD,k,c)2.(12)\zeta_k(i)=\sum_{c=1}^{C}\rho^{\mathrm{UE}}_{k,c}e^{j2\pi i f_{D,k,c}T}e^{-2(\pi T i\varsigma_{D,k,c})^2}. \tag*{(12)}

where ςD,k,c\varsigma_{D,k,c} represents the Doppler spread of said cluster, which can be related to the propagation geometry according to the following lemma.

Lemma 1. Let us consider a cluster seen by UE kk at an angle φk,c\varphi_{k,c} with an angle spread of ςUE,k,c\varsigma_{\mathrm{UE},k,c}. Assuming a Gaussian intra-cluster angular deviation, the cluster Doppler spread as seen by UE kk, denoted by ςD,k,c\varsigma_{D,k,c}, is defined as

ςD,k,c2=fˉD,k2[0.5(1+Ak,c2cos⁡(2ϵk,c))−Ak,ccos⁡2(ϵk,c)](13)\varsigma_{D,k,c}^{2}=\bar{f}_{D,k}^{2}\left[0.5\left(1+A_{k,c}^{2}\cos(2\epsilon_{k,c})\right)-A_{k,c}\cos^{2}(\epsilon_{k,c})\right] \tag*{(13)}

where Ak,c=e−ςUE,k,c2A_{k,c}=e^{-\varsigma_{\mathrm{UE},k,c}^{2}}, fˉD,k=fˉ∥νk∥/c0\bar{f}_{D,k}=\bar{f}\lVert\boldsymbol{\nu}_k\rVert/c_0, and ϵk,c=θk−φk,c\epsilon_{k,c}=\theta_k-\varphi_{k,c}.

Proof: Each ray scattered by cluster cc is seen by UE kk at an angle φk,c,n=φk,c+δUE,k,c,n\varphi_{k,c,n}=\varphi_{k,c}+\delta_{\mathrm{UE},k,c,n}, where δUE,k,c,n∼N(0,ςUE,k,c2)\delta_{\mathrm{UE},k,c,n}\sim\mathcal{N}(0,\varsigma_{\mathrm{UE},k,c}^{2}) represents the deviation from the nominal angle. Then, the Doppler shift of ray nn is fˉD,k,c,n=fˉD,kcos⁡(ϵk,c+δUE,k,c,n)\bar{f}_{D,k,c,n}=\bar{f}_{D,k}\cos(\epsilon_{k,c}+\delta_{\mathrm{UE},k,c,n}). Keeping in mind that E[cos⁡(δUE,k,c,n)]=exp⁡{−ςUE,k,c2/2}\mathbb{E}[\cos(\delta_{\mathrm{UE},k,c,n})]=\exp\{-\varsigma_{\mathrm{UE},k,c}^{2}/2\} and using the trigonometric identity cos⁡(a+b)=cos⁡(a)cos⁡(b)−sin⁡(a)sin⁡(b)\cos(a+b)=\cos(a)\cos(b)-\sin(a)\sin(b) we compute E[fˉD,k,c,n]=fˉD,kexp⁡{−ςUE,k,c2/2}cos⁡(ϵk,c)\mathbb{E}[\bar{f}_{D,k,c,n}]=\bar{f}_{D,k}\exp\{-\varsigma_{\mathrm{UE},k,c}^{2}/2\}\cos(\epsilon_{k,c}). Furthermore, by using the identity cos⁡2(a)=0.5(1+cos⁡(2a))\cos^{2}(a)=0.5(1+\cos(2a)), we are able to compute E[fˉD,k,c,n2]=0.5fˉD,k2(1+exp⁡{−2ςUE,k,c2}cos⁡(2ϵk,c))\mathbb{E}[\bar{f}_{D,k,c,n}^{2}]=0.5\bar{f}_{D,k}^{2}(1+\exp\{-2\varsigma_{\mathrm{UE},k,c}^{2}\}\cos(2\epsilon_{k,c})). The lemma follows by computing ςD,k,c2=E[fˉD,k,c,n2]−E[fˉD,k,c,n]2\varsigma_{D,k,c}^{2}=\mathbb{E}[\bar{f}_{D,k,c,n}^{2}]-\mathbb{E}[\bar{f}_{D,k,c,n}]^{2}

Channel estimation and spectral efficiency

Figure 1 shows the adopted pilot transmission pattern. Both UL and DL pilots are scheduled to provide to UEs and BS with estimated CSI for data detection and precoding. Each frame is composed of ΔUL\Delta_{\mathrm{UL}} UL data slots and ΔDL\Delta_{\mathrm{DL}} downlink ones, bringing the total number of data slots to Δ=ΔUL+ΔDL\Delta=\Delta_{\mathrm{UL}}+\Delta_{\mathrm{DL}}. The precoding matrices adopted during pilot transmission, denoted by W0,k,n\mathbf{W}_{0,k,n}, F0,n\mathbf{F}_{0,n}, are assumed to be fixed and known by both parties1. Let us now focus on the UL phase: each UE transmits its pilot matrix Zk∈CMUE×τp\boldsymbol{\mathcal{Z}}_k\in\mathbb{C}^{M_{\mathrm{UE}}\times\tau_p} in τp\tau_p subcarriers; these matrices are mutually orthogonal, i.e., ZkZk′H=τpULI⇔k=k′\boldsymbol{\mathcal{Z}}_k\boldsymbol{\mathcal{Z}}_{k'}^{\mathrm{H}}=\tau_p^{\mathrm{UL}}\mathbf{I}\Leftrightarrow k=k'. The considerations that follow apply to all blocks; we thus omit the index nn for brevity. Let us consider the arbitrary pilot slot ll, the observed pilot matrix is Yl=∑k=1KαkHk,lW0,k+Nl∈CMBS×τpUL\mathbf{Y}_l=\sum_{k=1}^{K}\alpha_k\mathbf{H}_{k,l}\mathbf{W}_{0,k}+\mathbf{N}_l\in\mathbb{C}^{M_{\mathrm{BS}}\times\tau_p^{\mathrm{UL}}}. After pilot de-spreading, the vectorized observation is defined as

Figure 1. Representation of the pilot and data symbol transmission structure in the time-frequency domain. The communication channel is estimated at the BS and at each UE thanks to the respective UL and DL pilots.

y~l,k=αkτpUL(W0,k⊗IMBS)hk,l+vec⁡(N~l)(14)\widetilde{\mathbf{y}}_{l,k}=\alpha_k\sqrt{\tau_p^{\mathrm{UL}}}(\mathbf{W}_{0,k}\otimes\mathbf{I}_{M_{\mathrm{BS}}})\mathbf{h}_{k,l}+\operatorname{vec}(\widetilde{\mathbf{N}}_l) \tag*{(14)}

The receiver uses pp pilot observations before the present one, bringing the total number of observations used for channel estimation to ptot=p+1p_{\mathrm{tot}}=p+1: We adopt the minimum mean square error (MMSE) channel estimation presented in [13], that is

h^k,i=αkτpULEk,iWptot,kHA‾k,i−1y~kptot(15)\widehat{\mathbf{h}}_{k,i}=\alpha_k\sqrt{\tau_p^{\mathrm{UL}}}\mathbf{E}_{k,i}\mathbf{W}_{p_{\mathrm{tot},k}}^{\mathrm{H}}\overline{\mathbf{A}}_{k,i}^{-1}\widetilde{\mathbf{y}}_{k}^{p_{\mathrm{tot}}} \tag*{(15)}

where

A‾k,i=αk2τpULWptot,kHMkWptot,k+σ2IptotMUEMBS,Wptot,k=Iptot⊗(W0,kT⊗IMBS),\overline{\mathbf{A}}_{k,i}=\alpha_k^{2}\tau_p^{\mathrm{UL}}\mathbf{W}_{p_{\mathrm{tot},k}}^{\mathrm{H}}\mathbf{M}_k\mathbf{W}_{p_{\mathrm{tot},k}}+\sigma^{2}\mathbf{I}_{p_{\mathrm{tot}}M_{\mathrm{UE}}M_{\mathrm{BS}}}, \qquad \mathbf{W}_{p_{\mathrm{tot},k}}=\mathbf{I}_{p_{\mathrm{tot}}}\otimes(\mathbf{W}_{0,k}^{\mathrm{T}}\otimes\mathbf{I}_{M_{\mathrm{BS}}}),
Ek,i=[ζ(i−l),…,ζ(i−l+p(Δ+1))]⊗CkUL,\mathbf{E}_{k,i}=[\zeta(i-l),\ldots,\zeta(i-l+p(\Delta+1))]\otimes\mathbf{C}_{k}^{\mathrm{UL}},
Mk=T([ζ(0),…,ζ(p(Δ+1))])⊗C^kUL,\mathbf{M}_k=\mathbf{T}([\zeta(0),\ldots,\zeta(p(\Delta+1))])\otimes\widehat{\mathbf{C}}_{k}^{\mathrm{UL}},
y~kptot=[y~k,l,…,y~k,l−p(Δ+1)]T.\widetilde{\mathbf{y}}_{k}^{p_{\mathrm{tot}}}=[\widetilde{\mathbf{y}}_{k,l},\ldots,\widetilde{\mathbf{y}}_{k,l-p(\Delta+1)}]^{\mathrm{T}}.

The channel estimate h^k,i\widehat{\mathbf{h}}_{k,i} is h^k,i∼CN(0,E‾k,i=αkτpULEk,iWptot,kHA‾k,i−1Wptot,kEk,iH)\widehat{\mathbf{h}}_{k,i}\sim\mathcal{CN}(0,\overline{\mathbf{E}}_{k,i}=\alpha_k\tau_p^{\mathrm{UL}}\mathbf{E}_{k,i}\mathbf{W}_{p_{\mathrm{tot},k}}^{\mathrm{H}}\overline{\mathbf{A}}_{k,i}^{-1}\mathbf{W}_{p_{\mathrm{tot},k}}\mathbf{E}_{k,i}^{\mathrm{H}}). On the other hand, The channel estimation error h~k,i=hk,i−h^k,i\widetilde{\mathbf{h}}_{k,i}=\mathbf{h}_{k,i}-\widehat{\mathbf{h}}_{k,i} is distributed as h~k,i∼CN(0,E‾k,i=CkUL−E‾k,i)\widetilde{\mathbf{h}}_{k,i}\sim\mathcal{CN}(0,\overline{\mathbf{E}}_{k,i}=\mathbf{C}_{k}^{\mathrm{UL}}-\overline{\mathbf{E}}_{k,i}). We now derive achievable UL and DL SE, assuming imperfect CSI and known precoders. We consider a generic subcarrier ν\nu belonging to the block nn. The BS adopts the arbitrary combining matrix Tk,i,n∈CMBS×MUE\mathbf{T}_{k,i,n}\in\mathbb{C}^{M_{\mathrm{BS}}\times M_{\mathrm{UE}}} for the detection

Λk,i,n=Tk,i,nH(∑l=1l≠kKαl2H^l,i,nWl,i,nWl,i,nHH^l,i,nH+∑l=1Kαl2Cov⁡[H~l,i,nWl,i,n]+σ2IMBS)Tk,i,n(16)\boldsymbol{\Lambda}_{k,i,n}=\mathbf{T}_{k,i,n}^{\mathrm{H}}\left(\sum_{\substack{l=1\\l\ne k}}^{K}\alpha_l^2\widehat{\mathbf{H}}_{l,i,n}\mathbf{W}_{l,i,n}\mathbf{W}_{l,i,n}^{\mathrm{H}}\widehat{\mathbf{H}}_{l,i,n}^{\mathrm{H}}+\sum_{l=1}^{K}\alpha_l^2\operatorname{Cov}\left[\widetilde{\mathbf{H}}_{l,i,n}\mathbf{W}_{l,i,n}\right]+\sigma^2\mathbf{I}_{M_{\mathrm{BS}}}\right)\mathbf{T}_{k,i,n} \tag*{(16)}
Γk,i,n=Qk,i,nH(∑l=1l≠kKαk2H^k,i,nDLFl,i,nFl,i,nH(H^k,i,nDL)H+αk2Cov⁡[H~k,i,nDLFi,n]+σ2IMUE)Qk,i,n(17)\boldsymbol{\Gamma}_{k,i,n}=\mathbf{Q}_{k,i,n}^{\mathrm{H}}\left(\sum_{\substack{l=1\\l\ne k}}^{K}\alpha_k^2\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\mathbf{F}_{l,i,n}\mathbf{F}_{l,i,n}^{\mathrm{H}}\left(\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\right)^{\mathrm{H}}+\alpha_k^2\operatorname{Cov}\left[\widetilde{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\mathbf{F}_{i,n}\right]+\sigma^2\mathbf{I}_{M_{\mathrm{UE}}}\right)\mathbf{Q}_{k,i,n} \tag*{(17)}
Tk,i,nMM=(∑k=1Kαk2(H^k,i,nWk,i,nWk,i,nHH^k,i,nH+Cov⁡[H~k,i,nWk,i,n])+σ2IMBS)−1αkH^k,i,nWk,i,n(18)\mathbf{T}_{k,i,n}^{\mathrm{MM}}=\left(\sum_{k=1}^{K}\alpha_k^2\left(\widehat{\mathbf{H}}_{k,i,n}\mathbf{W}_{k,i,n}\mathbf{W}_{k,i,n}^{\mathrm{H}}\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{H}}+\operatorname{Cov}\left[\widetilde{\mathbf{H}}_{k,i,n}\mathbf{W}_{k,i,n}\right]\right)+\sigma^2\mathbf{I}_{M_{\mathrm{BS}}}\right)^{-1}\alpha_k\widehat{\mathbf{H}}_{k,i,n}\mathbf{W}_{k,i,n} \tag*{(18)}
Qk,i,n=(αk2H^k,i,nDLFi,nFi,nH(H^k,i,nDL)H+Cov⁡[H~k,i,nDLFi,n]+σ2IMUE)−1αkH^k,i,nDLFi,n(19)\mathbf{Q}_{k,i,n}=\left(\alpha_k^2\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\mathbf{F}_{i,n}\mathbf{F}_{i,n}^{\mathrm{H}}\left(\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\right)^{\mathrm{H}}+\operatorname{Cov}\left[\widetilde{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\mathbf{F}_{i,n}\right]+\sigma^2\mathbf{I}_{M_{\mathrm{UE}}}\right)^{-1}\alpha_k\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\mathbf{F}_{i,n} \tag*{(19)}

of xk,i,v\mathbf{x}_{k,i,v}, while UE kk adopts the combining matrix Qk,i,n∈CMUE×MUE\mathbf{Q}_{k,i,n}\in\mathbb{C}^{M_{\mathrm{UE}}\times M_{\mathrm{UE}}} to detect xk,i,v\mathbf{x}_{k,i,v}. Assuming successive interference cancellation, UE’s kk UL SE at data symbol time ii is [6 Cor. 1]

SEk,i,vUL=E{log⁡2∣IMUE+αk2Hk,i,nHΛk,i,n−1Hk,i,n∣}.(20)\mathrm{SE}_{k,i,v}^{\mathrm{UL}}=\mathbb{E}\left\{\log_2\left|\mathbf{I}_{M_{\mathrm{UE}}}+\alpha_k^2\mathcal{H}_{k,i,n}^{\mathrm{H}}\boldsymbol{\Lambda}_{k,i,n}^{-1}\mathcal{H}_{k,i,n}\right|\right\}. \tag*{(20)}

where Hk,i,n=Tk,i,nHH^k,i,nWk,i,n\mathcal{H}_{k,i,n}=\mathbf{T}_{k,i,n}^{\mathrm{H}}\widehat{\mathbf{H}}_{k,i,n}\mathbf{W}_{k,i,n} and Λk,i,n\boldsymbol{\Lambda}_{k,i,n} is defined in (16). Here H~k,i,n\widetilde{\mathbf{H}}_{k,i,n} has been obtained by stacking h~k,i,n\widetilde{\mathbf{h}}_{k,i,n} into a matrix. The DL channel estimator is not explicitly reported for brevity, as it follows the same covariance-based approach of (15) and is reported in [11]. Assuming known BS precoders at the UEs, the DL SE can be modelled as

SEk,i,vDL=E{log⁡2∣IMUE+αk2(Hk,i,nDL)HΓk,i,n−1Hk,i,nDL∣}.(21)\mathrm{SE}_{k,i,v}^{\mathrm{DL}}=\mathbb{E}\left\{\log_2\left|\mathbf{I}_{M_{\mathrm{UE}}}+\alpha_k^2\left(\mathcal{H}_{k,i,n}^{\mathrm{DL}}\right)^{\mathrm{H}}\boldsymbol{\Gamma}_{k,i,n}^{-1}\mathcal{H}_{k,i,n}^{\mathrm{DL}}\right|\right\}. \tag*{(21)}

where Hk,i,nDL=Qk,i,nHH^k,i,nDLFk,i,n\mathcal{H}_{k,i,n}^{\mathrm{DL}}=\mathbf{Q}_{k,i,n}^{\mathrm{H}}\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}\mathbf{F}_{k,i,n} and Γk,i,n\boldsymbol{\Gamma}_{k,i,n} is defined in (17). Therefore, the average UL SE, and analogously the DL SE, over the frame, aggregated over the subcarriers and UEs, is defined as SEUL=∑k=1K∑i=1Δ∑v=1ΔULSEk,i,vUL/(Δ+2)\mathrm{SE}^{\mathrm{UL}}=\sum_{k=1}^{K}\sum_{i=1}^{\Delta}\sum_{v=1}^{\Delta^{\mathrm{UL}}}\mathrm{SE}_{k,i,v}^{\mathrm{UL}}/(\Delta+2).

Precoding strategy

The adopted precoding strategy differs between data slots, where CSI is available, and pilot slots, where the previous estimate has drifted too far from its actual value. In the absence of CSI, the BS implements eigen-beamforming, that is F0,n,k=[uBS,k,1,…,uBS,k,MUE]\mathbf{F}_{0,n,k}=[\mathbf{u}_{\mathrm{BS},k,1},\ldots,\mathbf{u}_{\mathrm{BS},k,M_{\mathrm{UE}}}], while W0,k,n=IMUE\mathbf{W}_{0,k,n}=\mathbf{I}_{M_{\mathrm{UE}}}. In the presence of CSI, the BS adopts MMSE precoding [14 Ch. 5] while the UEs precoders are Wk,i,n=Uk,i,n\mathbf{W}_{k,i,n}=\mathbf{U}_{k,i,n}, where Uk,i,n\mathbf{U}_{k,i,n} is the left singular matrix of H^k,i,nDL\widehat{\mathbf{H}}_{k,i,n}^{\mathrm{DL}}. Precoder[s] are normalized to ensure an equal power allocation between subcarriers and streams. We compare zero-forcing (ZF) and MMSE combining at the BS: the ZF combining matrix is Tk,i,nZF=Hi,nH\mathbf{T}_{k,i,n}^{\mathrm{ZF}}=\mathbf{H}_{i,n}^{\mathrm{H}}, with Hi,n=[α1H1,i,n,…,αKHK,i,n]\mathbf{H}_{i,n}=[\alpha_1\mathbf{H}_{1,i,n},\ldots,\alpha_K\mathbf{H}_{K,i,n}], while the MMSE combining matrix Tk,i,nMM\mathbf{T}_{k,i,n}^{\mathrm{MM}} is defined in (18). The UEs adopt the combining matrix Qk,i,n\mathbf{Q}_{k,i,n}, defined in (17).

Numerical results

The default value of the simulation parameters is reported in Table I. The effective radii are drawn independently from the log-normal distribution shown in Table I: rcr_c is clipped between 2 and 30 m to avoid unrealistically large radii. The resulting angle spreads are then also clipped between 1∘1^\circ and 10∘10^\circ. The number of clusters is divided into Cdom=⌈ηC⌉C_{\mathrm{dom}}=\lceil\eta C\rceil dominant clusters and Cback=C−CdomC_{\mathrm{back}}=C-C_{\mathrm{dom}} background one[s]; this allows us to control the propagation environment by

Parametervalue
BS and UE antennas (MBS,MUE)(M_{\mathrm{BS}}, M_{\mathrm{UE}})16, 4
UE positions (e1,e2)(e_1,e_2)300[cos⁡(±40∘),sin⁡(±40∘)]300[\cos(\pm40^\circ),\sin(\pm40^\circ)] m
UE pathloss (αk)(\alpha_k)3GPP street canyon model [15]
UE velocities (ν1,ν2)(\nu_1,\nu_2)10[cos⁡(0),sin⁡(0)]10[\cos(0),\sin(0)] m/s
Frequency blocks, (B)(B)2
Subc. per block (νcho)(\nu_{\mathrm{cho}})20
Carrier frequency, Symbol time (f‾,T)(\overline{f},T)4 GHz, 53 μ\mus
UL and DL data slots (ΔUL,ΔDL)(\Delta_{\mathrm{UL}},\Delta_{\mathrm{DL}})5
UL and DL pilot length (τPUL,τPDL)(\tau_{\mathrm{P}}^{\mathrm{UL}},\tau_{\mathrm{P}}^{\mathrm{DL}})20
BS and UEs total power (PBS,PUE)(P_{\mathrm{BS}},P_{\mathrm{UE}})30, 20 dBm
Number of clusters (C)(C)25
cc-th cluster effective radius (rc)(r_c)rc∼Lognorm⁡(log⁡(8),0.5)r_c\sim\operatorname{Lognorm}(\log(8),0.5)
RMS delay spread (DS)(DS)150 μ\mus
proportionality constant (τT)(\tau_T)2.4
Shadowing variance (στ2)(\sigma_\tau^2)3
UEs and BS noise variance (σBS2,σk2)(\sigma_{\mathrm{BS}}^2,\sigma_k^2)−150-150 dB

Table I. Simulation parameters default values

tuning η\eta. The dominant clusters are divided among the UEs as Ckdom=⌈Cdom/K⌉C_k^{\mathrm{dom}}=\lceil C_{\mathrm{dom}}/K\rceil and their position is randomly generated as pkdom=βk,cek+κk,cvk⊥\mathbf{p}_k^{\mathrm{dom}}=\beta_{k,c}\mathbf{e}_k+\kappa_{k,c}\mathbf{v}_k^{\perp}. Here βk,c∼U(0.35,0.65)\beta_{k,c}\sim\mathcal{U}(0.35,0.65), κk,c∼N(0,20 m)\kappa_{k,c}\sim\mathcal{N}(0,20\ \mathrm{m}) while vk⊥\mathbf{v}_k^{\perp} is a unit vector orthogonal to ek\mathbf{e}_k. On the other hand, the background clusters positions pcback=[pc,xback,pc,yback]T\mathbf{p}_c^{\mathrm{back}}=[p_{c,x}^{\mathrm{back}},p_{c,y}^{\mathrm{back}}]^{\mathrm{T}}, with pc,xback∼U(0,225)p_{c,x}^{\mathrm{back}}\sim\mathcal{U}(0,225), and pc,yback∼U(−300,300)p_{c,y}^{\mathrm{back}}\sim\mathcal{U}(-300,300). The proposed Weiscelberger model is denoted by “Geo”, while the Kronecker-separable baseline is labeled as “Kron”. Figure 2 shows the empirical Cumulative distribution function (CDF) of the covariance mismatch between the Kronecker separable model and the jointly correlated one: the solid curves show the nominal geometry while the dashed ones represent setups where the dominant-clusters centroid is shifted by 30∘30^\circ. We observe that the accuracy of a Kroneckerseparable approximation is strongly favored by a concentrated scattering environment (i.e., η=0.9\eta= 0.9), exhibiting a 50% probability of being below 25%. A dominant-cluster centroid shift makes the Kronecker separable approximation much less accurate and less dependent on η\eta, as the gap between the dotted curves shrinks substantially. This indicates that the Kronecker accuracy is favored by a strongly concentrated propagation environment where the clusters lie on the line between BS and UE. We now move to investigating the impact of a Kronecker-separable approximation on the achievable SE: Figure 3 shows the average UL and DL SEs evolution for increasing UE mobility in diffusive (i.e. η=0.1\eta= 0.1) and concentrated (i.e. η=0.9\eta= 0.9) scattering conditions. Figs 3a and 3b show the UL SE with MMSE and ZF BS combining, respectively, while Fig. 3c shows the DL SE. Increasing ∥ν∥\lVert\nu\rVert degrades all the SEs due to an increased aging rate of the channel. A concentrated scattering environment generates higher SEs: concentrated scattering geometries produce smaller Doppler spreads and their ζk(⋅)\zeta_k(\cdot) is less sensitive to UE mobility. The DL SE in Fig. 3c is more sensitive to ∥ν∥\lVert\nu\rVert than the UL, as the UL SE benefits from centralized BS-side detection, and shows a low-to-moderate Geo-Kron gap. On the contrary, the UL SE is less sensitive to UE mobility and exhibits a large Geo-Kron gap. Interestingly, with ZF BS combining, the Geo-Kron gap becomes smaller in concentrated propagation environments.

CDF of covariance mismatch

Figure 2. CDF of the covariance mismatch between the Kronecker separable model and the Jointly correlated one. The CDF is computed over 500 independent geometric setups.

Average spectral efficiency for uplink and downlink under different combining schemes and scattering conditions

Figure 3. Average SE in the UL, for both MMSE (a) and ZF (b) BS combining, and in the DL (c). We obtain the reported curves by averaging 20 Monte Carlo realizations of the propagation environment; for each environment, we average the SE over 500 channel and precoder realizations.

Conclusions

We proposed a geometry-aware channel model where the same clustered propagation environment determines both the Weichselberger eigen-mode coupling and the temporal correlation decay. Channel aging is now determined by the UE motion relative to the scattering geometry, rather than its velocity magnitude alone. Using explicit UL and DL pilots, we evaluated channel estimation and imperfect-CSI SE. The results show that Kronecker separability is an adequate approximation in highly concentrated propagation regimes, but can miss relevant eigenmode coupling in diffuse environments. Moreover, diffuse propagation induces stronger Doppler diversity, causing stronger channel aging and larger SE degradation at high UE velocities. This motivates eigenmode-coupled models for mobility-sensitive MIMO and ISAC systems.

References

  1. [1]S. Fodor, G. Fodor, D. Gürgünoğlu, and M. Telek, “Optimizing pilot spacing in MU-MIMO systems operating over aging channels,” IEEE Transactions on Communications, vol. 71, no. 6, pp. 3708–3720, 2023.arxiv.org/abs/2204.00213
  2. [2]K. T. Truong and R. W. Heath, “Effects of channel aging in massive MIMO systems,” Journal of Communications and Networks, vol. 15, no. 4, pp. 338–351, 2013.arxiv.org/abs/1305.6151
  3. [3]C.-X. Wang and M. Patzold, “Efficient simulation of multiple cross-correlated Rayleigh fading channels,” in 14th IEEE Proceedings on Personal, Indoor and Mobile Radio Communications, 2003. PIMRC 2003., vol. 2. IEEE, 2003, pp. 1526–1530.DOI
  4. [4]M. Bengtsson and P. Zetterberg, “Some notes on the Kronecker model,” EURASIP Journal on Wireless Communications and Networking, 2006.
  5. [5]W. Weichselberger, M. Herdin, H. Özcelik, and E. Bonek, “A stochastic mimo channel model with joint correlation of both link ends,” IEEE Transactions on wireless Communications, vol. 5, no. 1, pp. 90–100, 2006.DOI
  6. [6]Z. Wang, J. Zhang, B. Ai, C. Yuen, and M. Debbah, “Uplink performance of cell-free massive MIMO with multi-antenna users over jointly-correlated Rayleigh fading channels,” IEEE Transactions on Wireless Communications, vol. 21, no. 9, pp. 7391–7406, 2022.arxiv.org/abs/2110.04962
  7. [7]K. Dovelos, M. Matthaiou, H. Q. Ngo, and B. Bellalta, “Massive mimo with multi-antenna users under jointly correlated ricean fading,” in ICC 2020-2020 IEEE International Conference on Communications (ICC). IEEE, 2020, pp. 1–6.DOI
  8. [9]M. Series, “Guidelines for evaluation of radio interface technologies for imt-2020,” Report ITU, vol. 2512, no. 0, 2017.
  9. [10]R. Sinkhorn and P. Knopp, “Concerning nonnegative matrices and doubly stochastic matrices,” Pacific Journal of Mathematics, vol. 21, no. 2, pp. 343–348, 1967.
  10. [11]S. Rivetti, G. Fodor, E. Björnson, and M. Skoglund, “On the impact of channel aging and doppler-affected clutter on OFDM ISAC systems,” arXiv preprint arXiv:2601.05032, 2026.arxiv.org/abs/2601.05032
  11. [12]E. B. Kama, J. Kim, and E. Björnson, “Downlink pilots are essential for cell-free massive MIMO with multi-antenna users,” in 2024 IEEE Wireless Communications and Networking Conference (WCNC). IEEE, 2024, pp. 1–6.arxiv.org/abs/2404.18516
  12. [13]S. Rivetti, G. Fodor, E. Bjornson, and M. Skoglund, “Cram”er-rao bound analysis of bistatic isac under partial symbol knowledge and clutter,” arXiv preprint arXiv:2604.11426, 2026.arxiv.org/abs/2604.11426
  13. [14]Ö. T. Demir, E. Björnson, L. Sanguinetti et al., “Foundations of user-centric cell-free massive MIMO,” Foundations and Trends® in Signal Processing, vol. 14, no. 3-4, pp. 162–472, 2021.DOI
  14. [15]European Telecommunications Standards Institute (ETSI), “Study on channel model for frequencies from 0.5 to 100 GHz: 3GPP TR 38.901 version 15.0. 0 release 15,” 2018.

Paper details

vlnc:W_73egpkyf
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY-SA 4.0
Usage

views · 10

downloads · 1

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Not recorded for this paper

View statement
Looking for related papers…
Contents