Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Advance Computing and Microelectronics Unit (ACMU-Kolkata)


Ghosh, Sasthi Charan (ACMU-Kolkata; ISI)

Abstract (Summary of the Work)

In device to device (D2D) communication, two users residing in close proximity can directly communicate between them, through a common channel, without the need of a base station. A pair of D2D users forms a link and a channel needs to be allocated to it. The interference relationship among the active links at time t is modelled as an interference graph g(t). To establish interference-free communication, we have to assign a channel vector C(t) and a power vector corresponding to the active links such that the required signal to interference plus noise ratio (SINR) is satisfied for each link. Since channels are costly resources we have to minimize Y(t), the maximum channel used in C(t). Due to the movement of the D2D users, a channel allocated at time (t − 1) may create interference at time t. Hence a link may require to switch its channel to maintain its SINR. Channel switch itself produces delay and hence an additional overhead. Hence, to maintain quality of service, A(t), the total number of channel switches or perturbations in C(t) from C(t − 1), should also be minimized. Since each transmitter has a limited battery power, minimizing the total power P(t) used at time t, is also an objective of minimization. Note that if we allocate each link a different channel then each link has no interference from other links. In that case, A(t) is zero and each link can operate with the minimum power, resulting P(t) to be the minimum. But Y(t) is huge in that case. Thus Y(t)-A(t) and Y(t)-P(t) have natural trade-offs among them. Due the hardness, optimizing Y(t), A(t) and P(t) owing to their respective trade-offs is a challenging task. In this thesis, we developed a randomized algorithm as well as its parallel version which can find minimum Y(t) in expected polynomial time. To minimize A(t), we developed a centralized and a decentralized differential coloring technique as well as a random coloring technique. We calculated expected perturbations produced by each of them. To minimize a cost defined as a linear combination of Y(t) and A(t), we proposed a geometric prediction based and a graph union based approach and calculated the expected cost produced by them. Finally to minimize a cost defined as a linear combination of Y(t) and P(t), we proposed a randomized joint channel and power allocation technique and calculated the expected cost and energy efficiency produced by it theoretically as well as through simulations.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


Included in

Mathematics Commons