To meet its performance targets, the future 5G networks need to greatly optimize the Radio Access Networks (RANs), which connect the end users to the core network. In this thesis, we develop mathematical models to study three aspects of the operation of the RAN in modern wireless systems. The models are analyzed using the techniques borrowed mainly from queueing theory and stochastic control. Also, simulations are extensively used to gain further insights. First, we provide a detailed Markov model of the random access process in LTE. From this, we observe that the bottleneck in the signaling channel causes congestion in the access when a large number of M2M devices attempt to enter the network. Then, in the context of the so-called Heterogeneous networks (HetNets), we suggest dynamic load balancing schemes that alleviate this congestion and reduce the overall access delay. We then use flow-level models for elastic data traffic to study the problem of coordinating the activities of the neighboring base stations. We seek to minimize the flow-level delay when there are various classes of users. We classify the users based on their locations, or, in dynamic TDD systems, on the direction of service the network is providing to them. Using interacting queues and different operating policies of running such queues, we study the amount of gain the dynamic policies can provide over the static probabilistic policies. Our results show that simple dynamic policies can provide very good performance in the cases considered. Finally, we consider the problem of opportunistically scheduling the flows of users with time-varying channels taking into account the size of data they need to transfer. Using flow-level models in a system with homogeneous channels, we provide the optimal scheduling policy when there are no new job arrivals. We also suggest the method to implement such a policy in a time-slotted system. With heterogeneous channels, the problem is intractable for the flow-level techniques. Therefore, we utilize the framework of the restless-multi-armed-bandit (RMAB) problems employing the so-called Whittle index approach. The Whittle index approach, by relaxing the scheduling constraints, makes the problem separable, and thereby provides an exact solution to the modified problem. Our simulations suggest that when this solution is applied as a heuristic to the original problem, it gives good performance, even with dynamic job arrivals.
|Translated title of the contribution||Resource allocation in wireless access network : A queueing theoretic approach|
|Publication status||Published - 2016|
|MoE publication type||G5 Doctoral dissertation (article)|
- resource allocation
- wireless network
- queueing theory