Queueing theoretic models for wireless multicast systems and power control

As a part of the departmental Distinguished Lecture Series, the Department of Electronics and Communication Engineering at SRM University-AP is organising the first lecture on the topic “Queueing theoretic models for wireless multicast systems and power control via deep reinforcement learning”. Dr Vinod Sharma, Professor in Electrical Communication Engineering, IISc, Bangalore will be the keynote speaker of the session which will be held on July 29, 2021, at 4.00 pm.

We consider a wireless Base station (BS) transmitting on a downlink to multiple users. The BS has a finite number of files that the users can request. The users may have their own local caches. If a user requests a file and it is not in its cache, the request goes to the BS. This model is the practical scenario of downloading/streaming videos from Netflix. In order to reduce the transmission load on the downlink, we exploit the fact that the downlink is a broadcast channel and multiple users may be requesting the same file. This is accomplished by designing a novel multicast scheme in which all requests at the BS for the same file are merged together and are satisfied by a single transmission from the BS. This simple and natural scheme is shown to provide lower mean sojourn time compared to the currently popular coded caching schemes in literature and is much simpler to implement. We have also theoretically studied this queue and obtained approximate mean sojourn times. Next, we have shown that considering the channel fading can actually wipe out most of the benefits of our multicast scheme. However, we suggest a simple modification to the scheme which restores most of the gains. Furthermore, we show that adaptively changing the transmit power at the BS based on the instantaneous channel gains of the users can further significantly reduce the mean sojourn times. To get optimal power control, we use DQN, a Deep Reinforcement Learning algorithm. It is modified to ensure that the long term power constraint is satisfied and that it tracks the time-varying channel and request statistics.

About the Speaker:
Vinod Sharma completed his B.Tech Electrical Engineering from IIT Delhi, in 1978, PhD in Electrical and Computer Engineering, from Carnegie Mellon University, in 1984. He is Professor in Electrical Communication Engg., IISc, since March 2000. Prior to it, he was Associate Professor in Electrical Engineering, IISc Bangalore. He was Chairman of the Department of Electrical Communication Engineering from July 2007 to November 2011. He was Assistant Professor, Electrical and Communication Engg, Northeastern University; and Visiting Faculty, Electrical Engineering, University of California, Los Angeles. Dr Vinod’s research interests are in the area of Communication and Computer Networks, Wireless Communication, Sensor Networks, Queueing Theory, Information Theory and Statistical Estimation Theory.

Join this free informative webinar on July 29, 2021, from 4.00 pm onwards to gain research insights from the field expert.

Register here:https://srmap.zoom.us/webinar/register/WN_npqiEhMbSZebRzIj_VmRrA

Leave a Reply

Your email address will not be published. Required fields are marked *