Estrategias evolutivas para la adquisición de conocimiento en controladores borrosos temporales difuminados, aplicadas al encaminamiento adaptativo en redes de comunicaciones
- Juan Ramón Velasco Pérez Director
- Luis Magdalena Layos Co-director
Defence university: Universidad de Alcalá
Fecha de defensa: 01 October 2009
- Daniel Meziat Luna Chair
- Sancho Salcedo Sanz Secretary
- Juan Botía Blaya Committee member
- Joaquín Cañada Bago Committee member
- Gregorio Fernández Fernández Committee member
Type: Thesis
Abstract
In this thesis a methodology to improve the adaptive distributed routing performance is proposed. This procedure is based in the use of evolutionary genetic-fuzzy systems applied to the adaptive distributed routing, in packetswitched network communications. This proposed methodology can be divided into three component parts: 1. Development of Faded Temporal Fuzzy Logic Controllers and Hybrid Fuzzy Temporal Rules-based Controllers. 2. Development of Genetic Algorithm “Heading”. 3. Application of Evolutionary Genetic Fuzzy Systems to the adaptive distributed routing in packet-switched networks. After the identification of problems associated to the control with Fuzzy Logic Controllers, Temporal Fuzzy Logic Controllers and Fuzzy Temporal Rulesbased Controllers, to solve them, in this document, it is proposed the use of Faded Temporal Fuzzy Logic Controllers and Hybrid Fuzzy Temporal Rulesbased Controllers, which include the concept of temporal fading. For the last controllers, it is presented their news in: knowledge representation contained in the rules and knowledge bases and reasoning strategies. After the identification of problems associated with the use of Genetic Algorithm applied over these latest controllers, to solve them, in this document, it is proposed the use of Genetic Algorithm “Heading”, which produce an improvement on the “fitness” of the knowledge base obtained, and on the learning speed. These improvements are ascribed to force the appearance of a useful group of rules (temporal and non temporal) with the same antecedent, so that it is eliminated the random search. For this Genetic Algorithm it is presented their news in: structure of knowledge basis and knowledge adquisition. The use of a single metric for adaptive routing, in communication networks, is insufficient to reflect the actual state of the link. To solve this problem, it is proposed the use of two metric: the average link packets delay and the link packets jitter delay, measurements from the previous sampling intervals, to obtain a single metric: the output variable of a Fuzzy Logic Controller. The experimental results show that the insertion, in the routing process, of Evolutionary Temporal Fuzzy Logic Controllers or Hybrid Fuzzy Temporal Rules-based Controllers improve the network performance. This improvement, it is due to obtain metric values that are adapted to each different circumstances, to avoid the link congestion and high routing oscillation.