Espace membre

Cet espace est dédié aux étudiants, aux enseignants et au personnel administratif de l'école

Valider

Mot de passe oublié?

Agenda

27/09/2017

Doctorate thesis defense of Saloua Hendaoui




Doctorate thesis defense on September 27th 2017 at 10H00 ,in Amphi, Sup’Com 2.


Entitled :Resource allocation in LTE-A(Advanced Long Term Evolution)

Presented by :Saloua Hendaoui





Committee


President

Mrs. Sihem GUEMARA

Professor at Sup'Com, Tunisia

 

 

 

Reviewers :

Mr. Aref MEDDEB

Professor at ENISO, Tunisia

 

Mr. André-Luc Beylot

Professor ENSEEIHT/INPT University

 

Examiner :

Mr. Mohamed SIALA

Professor at Sup'Com, Tunisia

   

Thesis Director

Mr. Sami TABBANE

Professor at Sup'Com, Tunisia

 

Abstract


Long Term Evolution Advanced is a modern wireless communication network specified towards the fourth mobile generation. Real time applications are the first need for current users, especially with the exponential rise of the smart wireless equipments.


Packet scheduling is among the LTE-A features as it plays a crucial role in the LTE-A's Radio Resource Management (RRM) in order to enhance the system's capacity and to guarantee the variety of the QoS requirements of the evolving services. Packet schedulers are designed to intelligently share the radio resources between the connected users with quality of service guarantee. In each transmission time interval of 1ms, the scheduler computes the metrics of all the traffic pending in the evolved nodeB buffers. Then radio resources are shared depending on these metrics.


The literature of the LTE-A scheduling is very rich with schedulers aiming to guarantee the Quality of Service of the modern wireless application. Various scheduling classifications were given in the literature. In this thesis we give new schedulers classification which was not given before: Pure channel aware schedulers / schedulers based on prediction. For the first classification, metric computation is done based on information extracted from the reports exchanged between the User Equipments and the evolved NodeB . For the second classification, the schedulers compute metrics based on the prediction independently of the channel quality. For each class, we give our contributions in an aim to overcome the existing problems.


In our first contribution, we propose to overcome the problem of fix and exact budget delay. We give a new adaptive scheduler, which updates the delay budget with the wireless channel condition. The proposal has a low complexity cost, which is not the case of schedulers based on fuzzy logic methods. Performed simulations ensure that the proposed policy enhances the behavior of the classic schedulers. It guarantees bounded delays for the Real Time (RT) traffics. This contribution is used in our second proposal which aims to determine the estimated value of the packet reception time. The main improvement is that we take into account both scheduling metric and nodes status. Basically, we become able to determine the time necessary to lead to a successful packet reach. Simulations show that the proposition can make the LTE-A network more efficient by decreasing the delays for the RT traffic. Also, it did not affect the performance of the NRT (Non Real Time) flows. Then, we extend the reception time estimation proposal to overcome the problem of radio block wasting. Thisproposal permits to enhance the RT QoS. The main improvement is that we profit from the concept of pre-emption in LTE downlink scheduling to decrease the delays and to resolve the problem of radio blocks wasting. Preemption is done based on QoS aware decision along with fairness control between the RT and the NRT flows. The proposed scheduler has a low complexity cost. In fact it is mainly based on very simple mathematical formulas (summation) and a low complex algorithm. The results obtained in the state of the art of this thesis show that the spectral efficiency of the LTE-A network is under the targeted value. In fact, the existing QoS aware schedulers reduce the delay of the RT traffic at the cost of lower spectral efficiency. Hence, we give our proposal which aims to enhance the system spectral efficiency by profiting from the multi-user diversity. The proposal is QoS aware and at the same time it enhances the spectral efficiency of the cellular system. A trade-off between fairness and spectral efficiency is guaranteed. The proposition is approved with simulation. The findings prove that the proposition overcome the existing schedulers.


For the second class, we give a proposition which predict the bit rate in an aim to reduce the real time delay. It creates a highway inside the eNB in order to assign the available radio blocks (RBs) to the RT flow without leaving them waiting in the eNB buffers. Fairness between the RT and the NRT flows is guaranteed thanks to the proposed initialization and update algorithms. Prediction is adapted in the PBR (Predicted Bit-Rate) algorithm. Nevertheless, the exchanged reports are not totally ignored. In fact, the update algorithm tests the experienced QoS requirements. In this approach, the results show that the average delays as well as the PLR of the RT applications are not significant comparing to all other existed schedulers. In the following propositions, we predict the expected delay as well as the expected packet loss ratio. We propose an enhanced version of our already proposed scheduler (third proposal). The proposed policy succeeds to provide a good trade-off between the delay and the PLR for the RT traffic. In addition, it guarantees the fairness between the RT and the NRT traffic.


Keywords


LTE-A; Radio Resource Allocation; Scheduling; downlink; Real Time; Quality of service; Channel reporting; Prediction