Speaker: Hemant Saggar
Affiliation: Ph.D. Candidate
Via Zoom Only: https://ucla.zoom.us/j/345873626
Abstract: Today an innumerable number of devices use the wireless spectrum for communication, these include cellphones, WiFi devices, military radios, public safety radios, satellite phones etc. This crowding is limiting the experience of each device either through interference or by waiting for their turn to communicate. So, how do we allow the limited spectral resource to reliably scale to many more devices? This is possible through Concurrent Communication where multiple links share the spectrum and communicate simultaneously using multi-antenna techniques. One promising technique is Interference Alignment (IA), that has been shown to be Degrees-of-Freedom optimal under many conditions. Still, IA is sensitive to the accuracy of channel knowledge and its ability to achieve high throughput under time varying wireless conditions is unproven. We make progress towards understanding these limitations and provide viable solutions.
We study an IA system under different models of the time varying channel and derive expressions for the achieved rate over time and system throughput. Using these, we can arrive at the optimal duration of the data phase that maximizes throughput. We propose two strategies that help to counter the effects of the time varying channel. First, data aided receiver beam-tracking along with link adaptation gives sizeable improvement in the received SINR. Second, updating the transmit beams during data transmission short feedback pilots improves the alignment at receivers. In faster varying channels, we get a more stable achieved rate and in slower varying channels, we see additional throughput gains. The conclusion from this work is that an IA system must be trained more frequently than the channel coherence time to ensure high throughput and beam adaptation during the data phase gives significant robustness to the system.
Lastly, we present IA based Medium Access Control (MAC) protocols that outperform traditional protocols. Our concurrent Carrier Sense Multiple Access (CSMA) protocol based on beam-nulling is backward compatible with CSMA and is able to increase the sum throughput by 2 to 3x. We also show that IA outperforms optimal Time Division Multiple Access under time varying conditions. Hence a well-designed IA system can enable reliable concurrent communications in a wireless network.
Biography: Hemant Saggar is a Ph.D. candidate in the Electrical & Computer Engineering Department at UCLA. He is pursuing his Ph.D. at UCLA since fall 2014. Before, he was a research staff and student at IIT Delhi, India for 2013-14. He did his Bachelors and Masters in Electronics & Communication Engineering at IIT Roorkee, India from 2006-11 and secured the Institute Silver Medal. He worked at General Electric, India after graduation and was an intern at Qualcomm, San Diego in Summer 2016. He was awarded the Guru Krupa fellowship in Spring 2018 and Dean’s GSR fellowship in 2014. He likes to sing, play guitar, and explore buses and trains to travel. His research interests are Interference Alignment, MIMO techniques, adaptive filters, and neuroscience.
Date(s) - Apr 15, 2020
9:30 am - 11:00 am