Automation and Optimization of Multi-Layered Greenfield FTTH (Fiber-to-the-Home) Network Design

Abstract

Networks are core to the business of telecommunication providers and utility companies. However, the cost of network deployment can be very high as it involves digging up the roads for civil works to install the required network infrastructure, which drives the need for the optimal design of networks. Furthermore, network design requires the consideration of many complex factors in addition to the civil layer design such as network equipment dimensioning and positioning, detailed planning of cables and ducts. Typically, network designs are created manually by network planners. The quality of the produced designs depends highly on the planners' experience. Due to tight time constraints imposed on them, it is almost impossible to evaluate different planning scenarios and achieve an optimal design with the minimum cost. In general, existing optimization approaches to tackle this kind of problem can be classified into three major categories: meta-heuristics such as genetic algorithms, exact methods including mathematical programming, and heuristics. These approaches can produce an efficient network design that reduces the overall network deployment cost and cut down the design time, making it possible to evaluate different planning scenarios at once. However, most of them do not consider real networks as input and assume that the possible locations of network equipment are already provided. This dissertation addresses the problem of multi-layered FTTH network designs that includes the civil, duct and cable layers. The civil layer specifies which roads in a given area need to be dug up. The duct layer indicates the number of ducts that would be laid on a road to accommodate optical cables. The cable layer species the locations of the network equipment, the connectivity among them and the routes of different cables to be installed. Different layers have conflicting criteria making it difficult to optimize the overall deployment cost. For example, a design that is optimal in terms of cost in the cable layer may be far from optimal in the duct layer. To capture the specifications of the design of the distribution network, a mathematical programming model is formulated and its complexity is proven. The proposed research opts for a 3-stage optimization technique that explores the combination of different optimization techniques, such as linear programming and meta-heuristics, to solve the problem defined by the model. This dissertation proposes a multi-layer end-to-end approach to automate and optimize the design of cable and duct layers taking into account all the practical planning rules provided by an industrial partner. The method proposed is tested on real, practical networks. The proposed approach has been deployed for trial by the network design department of an industrial partner.

Date of Award	Jun 2016
Original language	American English
Supervisor	Kin Fai Poon (Supervisor)