Downlink Pilots in Cell-Free Massive MIMO Networks: A Boon or a Bane for URLLC?

Cell-free massive MIMO (CF-mMIMO) networks are a key enabler for beyond-5G and 6G systems, addressing the demands of ultra-reliable low-latency communication (URLLC). Supporting URLLC, characterized by short packets under strict latency and reliability constraints, poses challenges for CF-mMIMO. This paper extends finite-blocklength theory from cellular MIMO to scalable user-centric (UC) CF-mMIMO, exploring various downlink detection/decoding strategies. Random coding union bounds are derived to evaluate block error rates (BLER) for hardening-bound-based, pilot-based, and blind-based channel state information methods with centralized and distributed precoding, including distributed conjugate beamforming, local partial MMSE, and centralized improved partial MMSE. Our findings reveal that, in the short blocklength regime characteristic of URLLC scenarios, conventional blind CSI acquisition methods, as well as DL pilot-based strategies, underperform. To address this issue, an innovative MMSE-aided blind CSI acquisition strategy is proposed, which achieves superior performance with modest computational complexity. Numerical experiments show that, although centralized IP-MMSE precoders, characterized by rich channel hardening, enable near-optimal performance with simple hardening-based detection, the proposed MMSE-aided blind channel estimation approach emerges as the preferred detection/decoding strategy in nearly all considered scenarios, regardless of the precoding strategy, network load, or CF-mMIMO network infrastructure configuration.