Chapter Summary

Chapter 27 Summary: Open Problems

Key Points

• 1.

  The 3GPP channel model and XL-MIMO physics disagree. Current 5G NR channel models (TR 38.901) assume wide-sense stationarity across the array. Measurements from XL-MIMO campaigns at Lund, Eurecom, and Beijing routinely show per-user visibility regions covering only $40$-$60$ percent of the aperture. A combiner built on the WSS assumption loses $10\log_{10}(N_t/V_k)$ dB per user, yet no tractable parametric replacement has been standardized. Section 27.1 states the problem and the constraints any accepted solution must satisfy.

• 2.

  Centralized cell-free has an $\mathcal{O}(N_{\text{AP}}^3)$ wall. Scaling optimal MMSE processing to ultra-dense cell-free ($N_{\text{AP}} \gtrsim 10^3$ per km$^2$) is infeasible. Distributed alternatives (consensus MMSE, federated estimation, message passing) scale linearly in $N_{\text{AP}}$ at the cost of a performance gap that does not vanish with iteration count on realistic AP graphs. Whether there is a fundamental floor separating centralized from distributed performance remains the key open question of Section 27.2.

• 3.

  Full-duplex massive MIMO is 130 dB of work. Commercial FD radios need roughly $140$-$150$ dB of self-interference cancellation, achieved in a cascade of passive isolation, analog/RF cancellation, digital cancellation, and spatial nulling. Laboratory prototypes reach $\sim 125$-$130$ dB; commercial deployment is gated by phase-noise floors, PA nonlinearity, and network-wide duplex coordination. The massive array surplus $N_t - K_{DL}$ contributes $\sim 20$ dB of free spatial cancellation, closing the theoretical budget; engineering the last $10$-$20$ dB of margin is the open problem of Section 27.3.

• 4.

  Holographic MIMO buys DoF at the fourth power. The Pizzo-Marzetta-Sanguinetti theorem says a continuous aperture of size $L$ communicating over distance $d$ has $\mathcal{O}((L/L_F)^4)$ effective DoF in the radiative near field, with $L_F = \sqrt{\lambda\,d}$. This is a $L^2$ improvement over classical $\lambda/2$ arrays. Manufacturable surfaces realize only a fraction of the theoretical DoF because of unit-cell spacing, finite phase resolution, and mutual coupling. The open problem of Section 27.4 is to close that gap for practical frequencies and apertures.

• 5.

  RIS panels are passive APs. A cell-free deployment that includes passive RIS panels alongside active APs realizes a unified architecture in which the RIS contributes spatial DoF at nearly zero compute, zero fronthaul, and zero transmit power cost — but at a cascaded $1/(d_1^2 d_2^2)$ pathloss and a quadratic-in-$N_{\text{RIS}}$ pilot overhead. The joint optimization of AP precoders and RIS phase profiles is a non-convex problem with $\sim 10^5$ variables for a typical deployment, beyond real-time solvability. Section 27.5 catalogues the five sub-problems (scheduling, channel estimation, backhaul, joint beamforming, economics) that must be solved together.

• 6.

  The research agenda is not closing. Section 27.6 synthesizes the five themes into a concrete PhD-scale program and aligns it with the industry-engineering gaps that block commercial deployment (calibration, fronthaul, power-per-bit, beam management, mobility). The theoretical framework of Chapters 1-26 is largely in place; the gap between theory and product is where the next decade of MIMO research will be done.