Machine Learning based Precoder design in massive MIMO systems

Machine Learning based Precoder design in massive MIMO systems

Internship Description

Massive MIMO is an exciting area of 5G wireless research. For next-generation wireless data networks, it promises significant gains that offer the ability to accommodate more users at higher data rates with better reliability while consuming less power. The efficient precoding design at the massive MIMO base station is the basic requirement to fulfil the promises expected by this state-of-the-art technology. The precoding matrices need not only to be estimated at the beginning of the communication, but also require to be updated in every communication block due to the change in block fading channel. The training overhead to update the precoder design is one of the factors affecting the data rates in massive MIMO systems. As a first step, the project will aim to identify state-of-the-art precoder designs. The main aim would be to devise an efficient (complexity-wise) precoding update mechanism. Machine learning (ML) can play a role to predict the new precoding matrices efficiently for the ongoing communication, thus improving the spectral efficiency (due to less training overhead) as well as the computational complexity. The aim is to identify the most appropriate features to address the problem. Basically, history or training data for ML, that would be mapping the estimated channels to the precoding matrices, could be helpful to develop a robust feature vector for ongoing ML-based precoding update design. 

Deliverables/Expectations

1.  Literature survey of precoder design methods in the field of massive MIMO systems.
2. Understanding channel estimation and precoder design in massive MIMO communication systems.
3. Algorithms (MATLAB code) for precoder design in massive MIMO systems.
4. Algorithms (MATLAB code) for ML based precoder design in massive MIMO systems.
5. Final report summarizing and explaining all project work and reporting results of evaluation tests performed. 

Faculty Name

Tareq Al-Naffouri

Field of Study

Electrical Engineering