Maximizing energy efficiency in the vector precoded MU-MISO downlink by selective perturbation

Christos Masouros*, Mathini Sellathurai, Tharmalingam Ratnarajah

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract / Description of output

We propose an energy-efficient vector perturbation (VP) technique for the downlink of multiuser multiple-input-single-output (MU-MISO) systems. In contrast to conventional VP where the search for perturbation vectors involves all users' symbols, here, the perturbation is applied to a subset of the transmitted symbols. This, therefore, introduces a performance-complexity tradeoff, where the complexity is greatly reduced compared to VP by limiting the dimensions of the sphere search, at the expense of a performance penalty compared to VP. By changing the size of the subset of perturbed users, the aforementioned tradeoff can be controlled to maximize energy efficiency. We further propose three distinct criteria for selecting which users' symbols to perturb, each of which yields a different performance-complexity tradeoff. The presented analytical and simulation results show that partially perturbing the data provides a favorable tradeoff, particularly at low-power transmission where the power consumption associated with the signal processing becomes dominant. In fact, it is shown that diversity close to the one for conventional VP can be achieved at energy efficiency levels improved by up to 300% compared to VP.

Original languageEnglish
Article number6827257
Pages (from-to)4974-4984
Number of pages11
JournalIEEE Transactions on Wireless Communications
Issue number9
Publication statusPublished - 1 Sept 2014

Keywords / Materials (for Non-textual outputs)

  • complexity reduction
  • energy efficiency
  • multi-user MIMO
  • non-linear precoding.
  • Vector perturbation


Dive into the research topics of 'Maximizing energy efficiency in the vector precoded MU-MISO downlink by selective perturbation'. Together they form a unique fingerprint.

Cite this