Projecte llegit
Títol: Avaluació d'un mecanisme d'agrupació d'usuaris en xarxes Wi-Fi per la IoT utilitzant Intel·ligència Artificial
Director/a: LÓPEZ AGUILERA, ELENA
Departament: ENTEL
Títol: Avaluació d'un mecanisme d'agrupació d'usuaris en xarxes Wi-Fi per la IoT utilitzant Intel·ligència Artificial
Data inici oferta: 22-12-2020 Data finalització oferta: 22-07-2021
Estudis d'assignació del projecte:
- GR ENG SIS TELECOMUN
- GR ENG TELEMÀTICA
Tipus: Individual | |
Lloc de realització: EETAC | |
Segon director/a (UPC): GARCÍA VILLEGAS, EDUARD | |
Paraules clau: | |
WLAN, IoT, Machine Learning, 802.11ah | |
Descripció del contingut i pla d'activitats: | |
Dins de les xarxes d'àrea local sense fils (WLAN), l'especificació
IEEE 802.11ah consisteix en l'estàndard Wi-Fi per IoT (Internet de les Coses). Aquestes xarxes presenten un número elevat d'usuaris, i és necessari establir mecanismes per agrupar aquests usuaris en grups de menor número. En aquest projecte es treballarà amb mecanismes d'agrupació basats en tècniques d'intel·ligència artificial. Es treballarà amb mecanismes ja existents, i en la proposta de solucions alternatives, utilitzant eines software de simulació. |
|
Overview (resum en anglès): | |
The IEEE 802.11ah standard defines a Low-Power Wide-Area Network (LPWAN) solution for IoT networks, including multiple features that help solve some of the main problems of IoT networks, such as the need to save energy, covering great distances and managing a lot of simultaneous connections. One of these features is RAW (Restricted Access Window) mechanism, which has the objective of grouping stations to reduce collisions, improve performance and save energy. However, the standard does not define a method for grouping stations into RAW groups.
This project will use methods suggested by other authors, with the purpose of optimizing a genetic algorithm created by Python, until it is viable for real IoT networks. To be more precise, an already existing genetic algorithm from a previous study will be used as a starting point. A genetic algorithm was chosen as a solution for the grouping problem, due to its problem solving capacity involving the selection of the best individual out of a population that evolves during multiple generations. This main characteristic can be applied to our situation, if we understand the population as different group combinations of a series of stations, each of which represents an individual that can be a possible solution to the problem. The best individual will be chosen as the final solution to the grouping we are looking for. A series of optimizations will be performed with the purpose of executing the code in a device with specifications more commonly found in IoT networks. The first step consists in analyzing all the parameters concerning the genetic algorithm, which have not been yet optimized, to attempt to improve the results of the algorithm as much as possible and, at the same time, reduce the total computational time. Afterwards, a series of methods extracted from published studies done by other authors will be considered, with the purpose of optimizing the functions of the genetic algorithm. Finally, it will be tested in a device with worse specifications than the average computer, to prove the code¿s viability in a more constrained device. |