Wireless networks established by the operators combine several radio technologies and cover large areas allowing users to reach the operator's core network using network interfaces corresponding to different radio technologies. The user in such a heterogeneous wireless network must decide which access networks to use because this choice influences the quality of the data transmissions. The idea of using several radio technologies jointly to transfer the same data stream has made its way and has been realized by the multipath TCP protocol (MPTCP) in 2013.
In this thesis, on the one hand, we propose two models of access networks selection in heterogeneous wireless networks named UFMDP (Utility Function and Markov Decision Process) and IUMDP (Integrated user Utility in Markov Decision Process). These access network selection models take into account the joint utilities of the operator and the user. The utility of the operator uses an economic framework based on the concept of shadow price of the access node derived from Markov decision processes which allows to express dynamic user behaviours in a network (access frequencies, duration of connections and used access nodes). The utility of the user utilizes a utility function that expresses the satisfaction of the received QoS Metrics values over those requested.
On the other hand, we propose an implementation of MPTCP in a computer network simulator to allow tests on the proposals of improvement for the concrete usage of MPTCP. This implementation based on the RFC (Request for Comments) is used to propose a data flows scheduler to different MPTCP sub-sessions.
The simulations relate the comparisons of the proposed access network selection models and the conventional access network selection method based on the received signal strength (RSS). These simulations combine the access network selection methods with the single-path and multipath data transfers using the MPTCP protocol. The results of the simulations illustrate a significant increase in the revenue that the operator derives from the use of its networks with the proposed selection models as well as improvements on the Quality of Service (QoS) metric values for the users compared to the values given by the RSS method.
| Date | 17 May 2019 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Zbigniew Dziong (Supervisor) & Jean Charles Gregoire (Co-supervisor) |
|---|
Mah, F. (Author),
Dziong (Supervisor) & Gregoire (Co-supervisor),
17 May 2019Student thesis: Doctoral thesis › Doctorate in Engineering: Engineering