The application of genetic algorithms to programming of pavement maintenance activities at the network level is demonstrated. The opera- tional characteristics of the genetic algorithm technique and its rele- vance to solving the programming problem in a Pavement Management System (PMS) are discussed. The robust search capability of genetic algorithms enables them to effectively handle the highly constrained problem of pavement management activities programming, which has an extremely large solution space of astronomical scale. Examples are presented to highlight the versatility of genetic algorithms in accom- modating different objective function forms. This versatility makes the algorithms an effective tool for planning in PMS. It is also demonstrated that composite objective functions that combine two or more different objectives can be easily considered without having to reformulate the genetic algorithm computer program. Another useful feature of genetic algorithm solutions is the availability of near-optimal solutions besides the “best” solution. This has practical significance as it gives the users the flexibility to examine the suitability of each solution when practical constraints and factors not included in the optimization analysis are considered. The traditional methods of programming pavement management system (PMS) activities based on ranking methods or subjective priority rules do not guarantee an optimal or near-optimal utiliza- tion of available resources. This is because the number of pavement management activities required to be carried out at the network level is very large, and an optimization analysis is required to iden- tify a pavement management program that would achieve an opti- mal or near-optimal utilization of available resources. Although the conventional optimization techniques, such as lin- ear programming, nonlinear programming, integer programming and dynamic programming, have been suggested for use in pro- gramming of pavement management activities (1–4), few highway authorities have employed these techniques in their PMS. This is due to difficulties in formulating and modeling the problem and lengthy computation time, as well as uncertainty associated with the quality of the solution computed. The authors have recently demonstrated that genetic algorithms are a robust optimization tool that can be employed to obtain sufficiently good solutions to the programming problem within a practical time frame ( 5,6 ). In the present paper, the versatility of genetic algorithms (GAs) as a pro- gramming tool for pavement management is demonstrated. The technique is applied to provide further insights into the pavement management programming problem and to highlight several inter- esting issues and applications that are of significance to pavement management. TRANSPORTATION RESEARCH RECORD 1643 Paper No. 98-0019 1 PRACTICAL TOOL FOR OPTIMIZATION ANALYSIS Complexity of Activities Programming in PMS Programming of pavement management activities at the network level over a planning period of weeks or months (or years in the case of budget planning) is a complex problem. The complexity arises from the usually large number of pavement segments that have to be considered and the differences in pavement characteristics in terms of structural design, pavement age, maintenance history, and traffic-loading. Other problem parameters include types of pave- ment distress, levels of distress severity, and methods of pavement maintenance or repair. Together they form an extremely large prob- lem space, and what is commonly known as the “combinatorial explosion” of the feasible solution space. A simple example will illustrate the problem feature described in the preceding paragraph. Consider for illustration a very small road network with only 10 road segments. Assume that there are 4 pave- ment repair activities and 3 pavement distress severity levels, and the analysis is to select the maintenance activities to be performed weekly for a period of 12 weeks (approximately 3 months). This problem has altogether (10 × 4 × 3) = 120 decision variables, and the total number of possible solutions of the weekly program is equal to (12 + 1) 120 or 4.7 × 10 133 . It would take a modern day supercomputer many years to enumerate all of the possible solutions. In an actual pavement management programming problem, there are other complications, such as the time deterioration of pavements, and constraints of resources, such as budget, personnel, equipment, and materials. In a problem like this, it is practically impossible to identify the global optimal solution, and there are great difficulties in applying conventional optimization techniques to solve the problem. This is where GAs become an attractive tool to provide sufficiently good or near-optimal solutions for practical applications. Genetic Algorithms as Robust Pavement Management Programming Tool The theoretical basis of GAs for pavement management program- ming has been presented by the authors elsewhere (5–7 ); only a brief introduction is given in this section. GAs are a robust search technique formulated on the mechanics of natural selection and natural genetics (8,9). The analysis begins with a pool of randomly selected feasible solutions. Each of these solutions is represented by a string structure of cells (known as a genotype) containing coded values of the deci- sion variables of the problem. New solutions (known as offspring) are Analysis of Pavement Management Activities Programming by Genetic Algorithms T. F. FWA, W. T. CHAN, AND K. Z. HOQUE Centre for Transportation Research, Department of Civil Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Republic of Singapore.