Distributed Multiple Access for OFDMA Femtocells.

摘要

Demand for data services in cellular networks is growing exponentially, due to the proliferation of high-end, multimedia-enabled mobile devices. To meet this challenge, cellular operators are moving toward a heterogeneous network architecture consisting of macrocells for wide-area coverage and smaller cells such as femtocells for capacity boost in local hotspots. The primary problem in such heterogeneous networks is mitigating inter-cell interference especially in dense deployments of residential femtocells. Traditionally, interference in wide-area cellular networks has been studied from a multi-cell resource allocation perspective, where radio resources, e.g. power, bandwidth, are allocated to different cells to reduce interference. This approach generally assumes a fully loaded network (i.e. many simultaneous active users in a cell) where the system performance is insensitive to the activity of a single user. This assumption is not suitable for femtocells which are designed to serve very few users and thus lacks the presumed traffic aggregation.;In this thesis, we first discuss the infeasibility of applying the classical multi-cell resource allocation framework to the femtocell case. Then, we motivate the case of using distributed, random access protocols as apposed to centralized interference mitigation techniques in the context of LTE femtocells. For this, we employ queuing analysis to compare the performance of centralized resource allocation schemes represented by an ideal power control to the performance of a simple random access protocols. The results show that, with a high probability, a simple un-optimized random access protocol such as Aloha would perform much better than an optimized power controlled network in most of the unsaturated traffic scenarios. This result serves as a strong motivation for the design of OFDMA-aware random access protocols that utilize the inherent frequency diversity of OFDMA to improve the MAC performance.;Next, we analyze OFDMA-Aloha which is the simplest protocol that utilize the frequency-dimension in OFDMA to improve random access performance. The protocol attempts to reduce the packet retransmission time using collision resolution over the frequency domain by switching subchannels randomly after each collision. However, this comes at the expense of expanded time scale, or larger slot size due to lower channel rates. We showed that when the network is lightly loaded, the reduction in the collision rate outweighs the effect of expanded slot size thus in these situations OFDMA-Aloha enjoys smaller packet delays than the single channel Aloha.;Then, we propose the Exponential Backoff in Frequency (EBF) algorithm to address the case when multiple packets need to be transmitted over multiple subchannels simultaneously. Instead of spreading the packets uniformly over K subchannels as would be done in a multichannel variant of Aloha (K-Aloha), the EBF algorithm keeps packet transmissions clustered at few frequency branches using a synchronized binary tree branching process over the K subchannels. Collisions are handled by reducing the accessed bandwidth exponentially and using an access probability that is inversely proportional to the accessed bandwidth. Analysis of a lightly loaded network, shows that EBF enjoys considerably less packet delay compared to the basic K-Aloha protocol.;Finally, we propose the OFDM-based Reservation Random access (OFDM-RR) protocol that utilizes the frequency-dimension in a different way. In this protocol, reservation requests are transmitted on randomly selected subchannels and the indices of these subchannels are then used to create an implicit ordering among competing nodes for conflict resolution. Our analysis shows that beyond some critical value of the network load, OFDM-RR significantly improves the system throughput compared to the canonical Reservation-Aloha (R-Aloha) protocol.