Multi-cell joint processing and artificial intelligence solutions

One key characteristic of mobile communication systems is the overlap between radio cells, which becomes more prevalent with increasingly dense and heterogeneous networks. We investigate new signal processing techniques to manage this overlap to reduce interference, or to actively combine signals across multiple cells and antennas (massive MIMO) in order to boost performance.


By multi-cell joint processing, we mean two or more heterogeneous cells or devices (for instance small cells, picocells, femto-cells, mobile cells, or even involving satellites) sharing their antennas and information known at the network for cooperative transmission and reception. In theory, multi-cell joint processing has remarkable advantages including large spatial multiplexing and diversity gain as well as enhanced inter-cell interference cancelling capability. However, its practical implementation faces challenges such as signal processing scalability, imperfect channel estimation, timing and frequency synchronisation, and constrained backhaul.


Our primary objective is to tackle those challenges, and bring multi-cell joint processing closer to practice. We have made significant contributions, such as:

  • Several breakthrough multiuser MIMO technologies (MINT, MINT2, and A-QAM), which are able to offer near-optimal uplink or downlink transmission with signal processing complexity comparable to sub-optimal linear algorithms. Our technologies are scalable to the size of MIMO networks;
  • An advanced channel estimation scheme (pseudo-pilot) that is able to decompose a number of pilot-contaminated channels, and produce accurate estimations for each individual channel. Our scheme does not rely on angle diversity, and works for the general pilot contamination problem;
  • Real-complex hybrid modulated distributed MIMO or cooperative MIMO, which demonstrates remarkable macro-diversity gain and inter-cell interference cancelling capability.

Research focus

Our current research focuses on the use of deep learning technology to scale up multiuser-MIMO or low-cost massive MIMO solutions, with joint consideration of waveform and FEC coding optimisation, in order to deliver broadband and low-latency wireless access.

Get in contact

If you are interested in this research or have a query then please contact the project lead, Dr Yi Ma:

Contact us

Find us

5G Innovation Centre
University of Surrey