Beltaief, Student Nader (2019) Gestion des ressources radio par la programmation dynamique et par l'apprentissage par renforcement PRE - Research Project, ENSTA.



Thanks to DTM (Dual Transfer Mode) technology, different types of voice calls and data sessions can be established by terminals. In fact, a cell equipped with a DTM can transfer, with the same resources, voice calls and data sessions at full rate and half rate. This project aims to discuss the possibility of transforming optimally some of the full-rate uses, be they voice calls or data sessions, into half-rate uses to avoid overload. This control problem will be modeled by a Markovian decision process (MDP) by defining a suitable reward system. Subsequently, we will try to determine the optimal policy associated to this MDP using dynamic programming methods and evaluate the performance of this strategy and its ability to minimize the percentage of lost calls. In addition, being applicable to more sophisticated situations, unlike dynamic programming, we will try also to apply reinforcement learning methods in the search for an optimal policy, to test them and to know if it is useful to apply them on other more complex telecommunications problems.

Item Type:Thesis (PRE - Research Project)
Subjects:Mathematics and Applications
ID Code:7564
Deposited By:Nader Beltaief
Deposited On:09 juin 2021 15:00
Dernière modification:09 juin 2021 15:00

Repository Staff Only: item control page