Adjunct Professor

School of Information Technology and Engineering

Research

In general my research interests focus on cutting-edge technology for 4G and B4G wireless systems such as LTE, LTE-A and beyond.

SU-MIMO, MU-MIMO combined with OFDMA and capacity approaching codes, such as Turbo Codes and LDPC, are  technologies that are expected to boost spectral efficiency to a level where capacity will be able to meet data rate requirements on the short-to-medium horizon (until 2014-2015).

Due to the increased use of smart phones, electronic tablets, along with bandwidth-hungry applications that run on these devices such as video streaming, it is expected that data rate requirements will  be much greater than the capacity that will soon be offered with the use of the technologies mentioned above.

Work is underway (as in 3GPP Release 10 and beyond)  to come up with other enabling technologies that will increase the capacity even further. At the time of this writing (early 2011), potential candidate technologies include femtocells, relays and COMP (some times referred to as Multi-Cell MIMO). These emerging technologies impose a fundamental shift in thinking: now the multiple cells are viewed as cooperating rather interfering with each other. Hence a broader network view is adopted. This in turn creates immense challenges some of which relate to theoretical and analytical aspects of the problem at hand. 

Because of the complexity involved in analyzing the performance of these B4G systems we expect that dependence on computer simulations to validate future algorithms will increase. Since existing computer-based simulation platforms have limited computing power (mainly due to processor speed and the fact that most algorithms are written in a serial fashion) they will not be fast enough to obtain simulation results in a timely manner, i.e. hours vs weeks of simulation time.   

We are currently investigating the use of Graphical Processor Units (GPUs) to speed up simulation time of some common communication algorithms such FFT/IFFT, Turbo Decoders, LDPC decoders, etc. While GPUs have been extensively used in general purpose computing in other fields for the past decade their use in wireless applications have only surfaced within the last two years. So literature in this area of research is quite limited. The few papers published on this subject show extremely encouraging results where GPUs have been shown to offer up to 700x speed-up factor compared to conventional PC-based platforms when simulating the performance of a commonly used LDPC decoder. Some results showed that GPUs can even be used in some real-time applications.

GPUs are able to offer such supercomputing power because of their massively parallel architecture, offering up to 500+ cores in some NVIDIA GPUs. The main challenge is then how best to exploit this parallel architecture to speed-up the simulation time of existing and future wireless algorithms. Our target is to substantially reduce the simulation time of the most complex end-to-end 4G and B4G wireless systems from few days/weeks to few minutes or few hours at the most.

Posted in | Leave a comment