SCHEDULING OF PAVEMENT MAINTENANCE ACTIVITIES

Abstract

The focus of this study is on the optimal scheduling of pavement maintenance activities to reduce traffic delays at the network level using genetic algorithms. This thesis develops a new method of pavement maintenance scheduling at the road network level. Pavement maintenance management involves the coordination and control of a comprehensive set of activities in order to maintain pavements with the optimal use of available resources. Using the technique of genetic algorithms (GAs) a pavement maintenance scheduling methodology has been developed to minimize traffic delays subject to the constraints of maintenance operational requirements. Two hypothetical problems to schedule the maintenance activities of an urban road network have been examined with this technique. In the first problem, 20 roads in a network have been scheduled for three maintenance teams over a time period from 6 am on a weekday to 8 am the following day. The objective of the analysis was to arrive at a schedule that minimizes the total traffic delay subject to the following specific constraints: (a) any maintenance task assigned to a maintenance team must begin and end within the standby period of the maintenance team; (b) a maintenance team can perform only one task at a time; and (c) the total working hours of a maintenance team must be at least 6 hours and not more than IO hours. Travel time was considered as a part of the maintenance repair time. Integer coding has been used for the decision variables and the problem has been solved with two GA programs using different constraint handling methods.

In the second problem a road network having 100 roads was considered an maintenance activities were scheduled over a period of 5 days. The objective of the analysis was again to minimize the total traffic delay subject to all the constraint considered in the first problem. However, the travel time between two successive maintenance repairs was considered as a variable depending on the travel distance, road type and traffic volume. This example serves to demonstrate the potential of GAs in solving the pavement maintenance problem at network level.

The constraint handling techniques used in this study were the DRAM method and the penalty method. The DRAM method is a "decoder" and "repair" algorithm, in which the creation of invalid individuals is either avoided or repaired by a repair algorithm. In the penalty method the basic idea is to convert a constraint problem into an unconstrained one. Each of the genotypes is chosen from the search space and if a constraint is violated then a penalty is added to its evaluation function value. The DRAM method was found to produce better results than the penalty method and was adopted in this study.