Chapter Summary

Chapter 15 Summary

Key Points

• 1.

  Wireless channels are sparse. In the delay domain, wideband channels have $s \ll L$ active taps; in the angular domain, arrays see a few scattering clusters. The pilot observation $\mathbf{y} = \boldsymbol{\Phi}\mathbf{h} + \mathbf{w}$ is a compressed-sensing measurement. CS reduces pilot overhead from $M \geq L$ (full LS) to $M \sim s \log(L/s)$, i.e.\ pilots scale with the number of physical paths, not the ambient dimension.

• 2.

  Activity detection is sparse recovery. With $K_{\text{total}} \gg K_a$ potential users, assigning each a pilot signature $\boldsymbol{\phi}_k$ makes the activity vector $K_a$-sparse in $K_{\text{total}}$ dimensions. Pilot length scales as $M \sim K_a \log(K_{\text{total}}/K_a)/\text{SNR}$. The covariance-based detector of Fengler, Haghighatshoar, Jung, and Caire is consistent even when $K_a > M$ provided $N_r$ is large — a CommIT-driven result foundational for unsourced random access and 3GPP RedCap/ambient-IoT.

• 3.

  DOA estimation is sparse recovery on an angular grid. $\ell_{2,1}$ joint-sparsity beats MUSIC in snapshot-limited and coherent-source scenarios. Atomic-norm minimization achieves gridless super-resolution via an SDP characterisation, eliminating basis mismatch. This is the algorithmic substrate of integrated sensing-and-communication (ISAC) in 6G.

• 4.

  Structured sparsity sharpens recovery. Block sparsity ($\ell_{2,1}$) saves a $\log B$ factor per group; hierarchical sparsity (Wunder-Caire HiHTP) saves both $\log(G/K)$ and $\log(B/s)$ factors simultaneously. In FDD massive MIMO, joint sparsity across subcarriers (Haghighatshoar-Caire) exploits shared angular support to slash CSI-feedback overhead — the mathematical basis for Release-19 CSI feedback compression.

• 5.

  CS and classical estimation meet in the measurement regime. When $M \geq L$ we should use LS; when $M \ll L$ and the channel is compressible, CS dominates. The interesting regime $M \in (s \log(L/s), L)$ is exactly where real wireless systems live — and where all three research lines of this chapter (sparse CE, massive access, ISAC) operate.