No document available.
Abstract :
[en] The fifth-generation (5G) of wireless networks majorly supports three categories of services,
namely, enhanced mobile broadband (eMBB), ultra-reliable and low latency communications (URLLC), and massive machine-type communications. Every service has its own set
of requirements such as higher data rates, lower latency in packet delivery, high reliability,
and network energy-efficiency(EE) to support applications including ultra-high definition
(UHD) video streaming, virtual reality (VR), autonomous vehicles, vehicular communications, smart farming, and remote-surgery, respectively. The existing one-size-fits-all services
network model is not a viable option to support these services with stringent requirements.
Therefore, accommodation of these different services on the same physical network while
ensuring their distinct QoS requirements is a major challenge. To address this problem, a
new concept called network slicing (NS) has emerged as a promising solution for the dynamic
allocation of resources to wireless services with different QoS demands. The NS can be formed in both the radio access network (RAN) and core network (CN) parts. In this thesis,
we concentrate on RAN resources slicing, and more specifically on challenges that reside in
the assignment of limited radio resources to manage the distinct traffic demands occurring
from a wide variety of users belonging to heterogeneous services. Specifically, we address the
RAN optimization method for joint allocation of time, frequency, and space resources to the
eMBB, URLLC, and mMTC users according to their traffic demands (i.e., queue status).
Our work in this thesis can be broadly classified into three parts based on the objective
function considered in the resource optimization problem: 1) sum-rate maximization, 2)
power minimization, 3) EE maximization.
In the first part of this thesis, we address an adaptive modulation coding (AMC) based
resource allocation problem for dynamic multiplexing of URLLC and eMBB users on the
shared resources of the OFDMA-based wireless downlink (DL) network. Specifically, we
formulate the resource blocks (RBs) allocation problem as a sum-rate maximization problem subject to the minimum data rate constraint, the latency-related constraint, orthogonality,
and reliability constraints. Furthermore, to allocate RBs and transmit power jointly to the
users, we also formulate the AMC-based optimization problem to maximize the sum good-put
of eMBB users subject to URLLC and eMBB users’ QoS constraints. Importantly, in this
problem formulation, we consider a probabilistic constraint to incorporate CSI imperfections
and a short-packet communication model for URLLC service.
In the second part of this thesis, we formulate the joint RBs and transmit power allocation problem to minimize the transmit power consumption at the BS while guaranteeing
the QoS constraints of eMBB, URLLC, and mMTC users and probabilistic constraint to incorporate CSI imperfection, respectively. Importantly, we consider mixed-numerology-based
RB grid models to the users according to their queue status/traffic demand for satisfying
their stringent requirements. Furthermore, different slicing strategies such as slice-aware
(SA) and slice-isolation (SI) resource assignment mechanisms are considered for the efficient
co-existence of URLLC, mMTC, and eMBB services on the same RAN infrastructure.
The resulting problems in the first and second parts of the thesis are mixed-integer non-linear programming problems (MINLPs) which are intractable to solve. To provide solutions
to these problems, we first transform the problems into more tractable using the AMC approximation functions, probabilistic to non-probabilistic conversion functions, Big-M theory,
and difference of convex (DC) programming. Later, these transformed problems are solved
using the successive convex approximation (SCA) based iterative algorithm. Our simulation
results illustrate the performance of our proposed method compared to the baseline methods.
Also, the simulation results show the effectiveness of the mixed-numerology-based RB grid
model over the fixed numerology grid model and the performance of SA and SI resource slicing strategies in terms of achievable data rates, packet delivery latencies, and queue status,
respectively.
In the third part of this thesis, different from the first two parts, for the joint assignment of
beams, RBs, and transmit power to eMBB and URLLC users, we formulate an EE maximization problem by considering resources scheduling, orthogonality, power-related constraints,
and QoS constraints for different services. The resulting mixed-integer non-linear fractional
programming problem (MINLFP) is intractable and difficult to solve. To provide a feasible
solution, we first transform the formulated problem into a tractable form using fractional
programming theory, approximation functions and later utilize the Dinklebach iterative algorithm, DC programming, and SCA to solve it. Finally, we compare the performance of the
proposed method against baseline schemes through simulation results. In particular, we show
the performance of RAN slicing mechanisms with the mixed and fixed-numerology-based RBs
grid models in terms of achievable EE, packet latencies, data rates, total sum-rate, and computational complexity. The proposed algorithm outperforms the baseline schemes in terms
of achieving higher data rates for eMBB users. The results also show the trade-off between
the total achievable sum rate and EE of the network. The proposed method with mixed
numerology grid delivers 20% of higher URLLC packets within 1 ms of latency. Besides, it
achieves the lowest computation time than that with the fixed numerology grid.