arXiv subjects

Atri Mukhopadhyay

Publications and source records attributed to Atri Mukhopadhyay.

3 recordsLinked to original sources

arXiv2021

Learning Aided Auctioning based Spectrum Access System in a Wireless Optical Network

This paper focusses on Service Level Agreement (SLA) based end-to-end Quality of Service (QoS) maintenance across a wireless optical integrated network. We use long term evolution (LTE) based spectrum access system (SAS) in the wireless network and the optical network is comprised of an Ethernet Passive Optical Network (EPON). The proposal targets a learning-based intelligent SAS where opportunistic allocation of any available bandwidth is done after meeting the SLA requirements. Such an opportunistic allocation is particularly beneficial for nomadic users with varying QoS requirements. The opportunistic allocation is carried out with the help of Vickrey-Clarke-Groves (VCG) auction. The proposal allows the users of the integrated network to decide the payment they want to make in order to opportunistically avail bandwidth. Learning automata is used for the users to intelligently converge to the optimal payment value based on the network load. The payment made by the users is later used by the optical network units of the EPON to prepare the bids for the auction. The proposal has been verified through extensive simulations.

cs.NI↗

arXiv2020

Low Complexity Fair Scheduling in LTE/LTE-A Uplink Involving Multiple Traffic Classes

The bulk of the research on Long Term Evolution/Long Term Evolution-Advanced packet scheduling is concentrated in the downlink and the uplink is comparatively less explored. In up-link, channel aware scheduling with throughput maximization has been widely studied while considering an infinitely back-logged buffer model, which makes the investigations unrealistic. Therefore, we propose an optimal uplink packet scheduling pro-cedure with realistic traffic sources. Firstly, we advocate a joint channel and buffer aware algorithm, which maximizes the actual transmitted bit-count. Thereafter, we introduce delay constraints in our algorithm to support real-time traffic. We further enhance our algorithm by incorporating the varied delay and throughput requirements demanded by mixed traffic classes. Finally, we in-troduce priority flipping to minimize bandwidth starvation of lower priority traffic in presence of higher percentage of high priority traffic. We observe that a delay constraint may render the optimization-based proposals infeasible. Therefore, to avoid infeasibility, we replace the delay constraint with delay outage minimization (DOM). DOM aims at minimizing the packet drop due to delay violation. Moreover, DOM also helps in reducing the problems to a well-known assignment problem, which can be solved by applying the Hungarian algorithm. Hence, our approach delivers an optimal allocation with low computational complexity.

cs.NI↗

arXiv2011

VoIP Call Optimization in Diverse Network Scenarios Using Learning Based State-Space Search Technique

A VoIP based call has stringent QoS requirements with respect to delay, jitter, loss, MOS and R-Factor. Various QoS mechanisms implemented to satisfy these requirements must be adaptive under diverse network scenarios and applied in proper sequence, otherwise they may conflict with each other. The objective of this paper is to address the problem of adaptive QoS maintenance and sequential execution of available QoS implementation mechanisms with respect to VoIP under varying network conditions. In this paper, we generalize this problem as state-space problem and solve it. Firstly, we map the problem of QoS optimization into state-space domain and apply incremental heuristic search. We implement the proposed algorithm under various network and user scenarios in a VoIP test-bed for QoS enhancement. Then learning strategy is implemented for refinement of knowledge base to improve the performance of call quality over time. Finally, we discuss the advantages and uniqueness of our approach.

cs.NI↗