| The effective maintenance of transportation infrastructure is essential for the continued development of nations engaged in commerce. Prompt and low cost maintenance is a primary concern of highway agencies increasingly seeking improved methods of inspection monitoring and maintenance, to maximize the life-cycle of their infrastructure, particularly roads, bridges, guardrails, signs and luminaries. Highway maintenance agencies have the continual challenge of inspecting their expanding highway networks, which deteriorate due to heavy vehicle axle loads, accidents, weather and other causal factors. Earlier research focused on infrastructure maintenance primarily on pavements and bridges due to their capital-intensive nature. However, there has been relatively little attention given to roadside features, such as guardrails, signs, and luminaries. Roadside features usually number into the hundreds of thousands, making the maintenance process very complex, and thus, require an automated procedure that would provide the benefit of optimal inspections and consequently, the efficient allocation of maintenance resources.; The objective of this dissertation is to develop two interrelated models, one for obtaining a minimal inspection travel schedule of roadside features, and another for determining the optimal Maintenance Rehabilitation and Reconstruction (MR&R) actions for roadside features over a given planning horizon, subject to budgetary constraints.; The first Model is formulated as a mixed integer-linear program to derive minimal inspection routes for a given road network. The developed formulation is an extension of the benchmark Capacitated Arc Routing Problem (CARP), to produce optimal travel routes for street networks that is efficiently traversed for inspection purposes. The second Model is formulated as a sequential decision process, with the objective of optimizing needed MR&R activities for a given highway network, over a planning horizon. It is assumed that the model exhibits the Markovian property due to the memory-less nature of roadside feature conditions over successive discrete time intervals.; The Genetic Algorithms procedure is applied to both Models by creating an initial population of solutions, and applying crossover and mutation operators to derive optimality. The first Model is similar in nature to the CARP and for the second Model roadside feature conditions are considered to be available exogenously to be used in the conduct of maintenance policies. Example problems are demonstrated to show the applicability of both models to real-world situations. This dissertation concludes with directions for future work for improved real-world practical application.; Keywords. Roadside Features, Highway Maintenance Scheduling, Capacitated Arc Routing Problem, Markov Decision Process, Roadside Features Life-Cycle, Genetic Algorithms. |
