Distributed Bi-Directional Training of Nonlinear Precoders and Receivers in Cellular Networks

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Joint optimization of nonlinear precoders and receive filters is studied for both the uplink and downlink in a cellular system. For the uplink, the base transceiver station (BTS) receiver implements successive interference cancellation, and for the downlink, the BTS station pre-compensates for the interference with Tomlinson-Harashima precoding (THP). Convergence of alternating optimization of receivers and transmitters in a single cell is established when filters are updated according to a minimum mean squared error (MMSE) criterion, subject to appropriate power constraints. Adaptive algorithms are then introduced for updating the precoders and receivers in a distributed manner without channel state information, assuming time-division duplex transmissions with channel reciprocity. Instead of estimating the channels, the filters are directly estimated according to a least squares criterion via bi-directional training: Uplink pilots are used to update the feedforward and feedback filters, which are then used as interference pre-compensation filters for downlink training of the mobile receivers. Numerical results show that nonlinear filters can provide substantial gains relative to linear filters with limited forward-backward iterations.

Original languageEnglish (US)
Article number7120181
Pages (from-to)5597-5608
Number of pages12
JournalIEEE Transactions on Signal Processing
Issue number21
StatePublished - Nov 1 2015


  • Bi-directional training
  • THP
  • distributed optimization
  • interference cancellation
  • interference pre-compensation
  • non-linear precoder
  • non-linear receive filter

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Distributed Bi-Directional Training of Nonlinear Precoders and Receivers in Cellular Networks'. Together they form a unique fingerprint.

Cite this