OPTIMIZATION IN PAVEMENT MANAGEMENT AT NETWORK LEVEL USING EVOLUTIONARY ALGORITHMS

Abstract

This thesis illustrates the applicability of evolutionary techniques to pavement management at network level. Pavement management involves the coordination and control of a comprehensive set of activities in order to maintain pavements with the optimal use of available resources. In the current study, a literature review of pavement management and an overview of evolutionary algorithms are presented. Evolutionary algorithms are optimization techniques that mimic the natural process of Darwinian evolution. The effectiveness of evolutionary algorithms as an optimization technique is then demonstrated through two such algorithms, EPAVE and EPA VE-INT. EPAVE was used to successfully locate a known optimum of a simple application problem and EPAVE-INT was used as a benchmark test to compare the efficiency of evolutionary algorithms with that of integer programming on a more complex combinatorial problem. Based on the principles of evolutionary algorithms, a computer model, PAVENET, was then formulated to solve a pavement maintenance planning problem at network level. Analyses are conducted to show the characteristics of important operating parameters of the PA VENET programme. Examples of maintenance planning of road networks are presented to analyze the effects of pavement parameters, maintenance policy parameters and resource parameters. An improved version of PAVENET, known as PAVENET-R, was subsequently created to handle the more complex problem of maintenance-rehabilitation tradeoffs. There are more constraints involved and the issue of constraint handling was discussed. The advantages and disadvantages of methods such as the penalty method and the decoder and repair methods for pavement management problems were described. PAVENET-R employed a different approach of using suitably chosen genetic representations and genetic operators to handle constraints. This is presented through an application problem. The ability of evolutionary algorithms to handle rehabilitation and maintenance tradeoffs augurs well for their use in pavement management systems. The concluding chapter recapitulates the essential points of the study and recommend areas for further research.