Cooperative transmission with distributed CSIT: fundamental limits and a new look at the log-det formula
The availability of accurate and, most importantly, shared channel state information at the transmitter (CSIT) is one of the key factors that enable transmitters cooperation in decentralized wireless systems. However, in some cases, channel information may not be easily or perfectly shared among the transmitters, thus limiting their coordination capabilities. In this paper we shed some light on the fundamental limits of networks with cooperating transmitters impaired by a general distributed CSITassumption. To this end, we consider a memory-less multiple-access channel with common message, and with noisy causal CSIT and noisy channel state information at the receiver (CSIR). Perhaps surprisingly, and in contrast to the same setting in absence of common message, we show that distributed precoding based on current CSIT only (namely, a Shannon strategy) achieves the sum-rate capacity of this channel, for every degree of CSIT and CSIR. We then illustrate this result in a practically relevant cooperative MIMO Gaussian channel with fading working in frequency division duplexing (FDD) mode. This setting allows us to revisit the celebrated log-det formula by showing that, in distributed settings, the classical approach of limiting the number of Gaussian streams to the maximum number of spatial degrees-of-freedom of the system may be suboptimal.
Lorenzo Miretti received the BSc and MSc degrees in Telecommunication Engineering from Politecnico di Torino in 2015 and 2018 respectively, both cum laude. From July 2017 to December 2017, he was a student researcher at Fraunhofer HHI Institute, Berlin, where he authored several publications on the topic of channel acquisition for Massive MIMO systems. In February 2018 he joined as a PhD student the Communication Systems Department of EURECOM, Sophia Antipolis, working on new paradigms for mobile networks based on recasting devices as distributed computational nodes, under the supervision of David Gesbert. His current research interests lie in the area of distributed optimization for communication networks, multi-antenna and multi-agent signal processing, and multi-user information theory.