Stochastic Game Analysis and Probabilistic Approach towards the Design of Random Access Mechanisms

Abstract

This thesis deals with the channel access problem in a wireless network using analytical approaches based on game theory and Markov chain theory. The goal is to improve the performance of the network. Particular emphasis is placed on channel access for unicast traffic and multicast traffic. Two random access mechanisms are studied in this thesis, namely, “Slotted ALOHA Enhanced by ZigZag Decoding (SAZD)” and “Multicast Collision Prevention (MCP)”. SAZD is an enhancement of the Slotted ALOHA random access mechanism by ZigZag decoding technique, while MCP is a multicast mechanism designed for high-performance video communications. MCP belongs to the same family of random access mechanisms as SAZD, and it is characterized by the functionality of channel sensing, implemented by the CSMA (Carrier-Sense Multiple Access) mechanism. Moreover, it is based on the DCF (Distributed Coordination Function) mechanism, which exploits the BEB (Binary Exponential Backoff) scheme to access the wireless channel. The thesis is composed of two parts. The first part is devoted to our contributions to the SAZD mechanism, and the second is devoted to our contributions to the MCP mechanism. For the SAZD mechanism, we initially proposed a cooperative game model to model a wireless network scenario where users (i.e., players) transmit data according to this mechanism. The proposed model takes into consideration the following constraints: maximizing throughput, minimizing delay, and optimizing the throughput-delay trade-off. Next, we introduce two cooperative game pricing strategies to control user behavior and further optimize the trade-off between throughput and delay. Then, we study the previous scenario using this time a non-cooperative game. Thus, we consider selfish users who share the same transmission channel. Our analysis shows that users, in this case, transmit using a very high probability. This aggressive behavior leads to a dramatic situation where the wireless network fails. To address this issue, we propose introducing a transmission cost to mitigate the aggressive behavior of users. Thus, we can decrease the level of the selfishness of users without adding a central coordination unit, which allows us to have a significant improvement in network performance. Afterward, we propose to look for the optimal cost which allows for transforming the non-cooperative game into a cooperative game like the one studied in the first chapter. We then extend the two previous models into a general game model in which cooperative and non-cooperative users coexist. This model generalizes several models that consider either the concept of cooperation or non-cooperation to model a wireless network system. For the MCP mechanism, we first propose a Markov model that models its operation in the case of saturated traffic. This model takes into account the existence of other types of traffic, such as unicast traffic, and it takes into account the backoff freezing if the channel is busy. Our goal is to achieve reliability of over 99.999% for multicast traffic. All the results found are validated by extensive simulations. We also study this model in the case of unsaturated traffic, and we evaluate different parameters of the Quality of Service (QoS) and the Quality of Experience (QoE) of the network based on the MCP mechanism. Finally, we propose a new approach to evaluate the energy consumption in the network. We estimate the energy consumed between two successive transmissions for unicast traffic and multicast traffic.