Contribution to the traffc engineering in wireless mesh networks
- Autores: Juan Pablo Astudillo León
- Directores de la Tesis: Luis J.de la Cruz Llopis (dir. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2020
- Idioma: español
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Nowadays, we live in a modern society in which people and devices are interconnected anywhere and anytime. Under this premise, both the infrastructure and the services offered have evolved and diversified in a drastic way. In fact, many of these services are transported in decentralized networks. Among them, Wireless mesh networks are decentralized networks that have been widely studied in different research areas such as community networks, public safety and surveillance. Wireless mesh networks have been also studied and evaluated in the Smart Grid scenario. Smart Grids are a new paradigm in which the electricity network is no longer focused only on the generation, distribution and transport of electricity to subscribers. Now, it is a robust network that includes a data communication network. The associated data network is divided in different subnetworks. This thesis is mainly focused on the improvement of the performance of one of those subnetworks, the so-called Smart Grid Neighborhood Area Network. Several applications are transmitted between the users and the control center. In general, upstream communication involves tasks such as meter reading, billing data or electricity consumption, while downstream communication allows the smart grid to take actions in different network situations such as power peaks or emergency situations. In the first part, the work is focused on improving the routing mechanism. To do this, a multipath routing mechanism is proposed, where the traffics that are most important are transmitted over the best communication links. In order to improve even more the benefits obtained, a multichannel scheme is proposed to separate both control traffic and data traffic, and use the less congested channels to transmit the most priority traffic types.Smart Grids offer many services and some of them are very demanding in terms of QoS. Besides, infrastructure failures, attacks and high congestion situations can greatly reduce the network performance. Therefore, the network must be able to offer a minimum QoS to the most priority applications handling some traffic control techniques. With this goal in mind, in this thesis some congestion control mechanisms are also proposed. In the first of these mechanisms, the decision of whether a packet should be retransmitted or not is made in a distributed and independent way by each one of the network nodes, depending on the network conditions that the node itself is observing. This mechanism considers again the existence of traffics with different priorities, so that, less priority traffic has a higher probability of being discarded. Furthermore, an emergency system is coupled to the congestion control mechanism. With this strategy, the NAN is able to take global actions (in a short time) to face anomalous situations.In a Smart grid scenario, the nodes are static and each of them transmits upstream data flows to the data concentrator. Therefore, depending on their geographical location, some nodes may be more favored than others. Besides, some nodes can monopolize the network resources if they are not regulated. For this reason, in this thesis another distributed solution is proposed that runs in each node. The objective here is to provide a fair distribution of network resources regardless of the geographical position and the transmission rate. The last contribution is focused on the application of machine learning techniques to obtain again a better performance of the data networks under study. In this sense, a new congestion control mechanism is proposed, which, like the previous ones, provides different quality of service to data flows with different priorities. For this, a complete framework is proposed, including the generation, preprocessing and evaluation of the data necessary for the training of the machine learning algorithms that will be used. The proposal is also implemented and evaluated in the Smart Grid NANs environment
