The use of genetic algorithms in Road Maintenance Optimisation. Fritz Jooste, Lonrix

The solution to the problem of finding an optimal set of road maintenance strategies under a multi-year budget constraint is exceedingly complex. Over the past few decades many different approaches to this problem have been implemented – these include linear programming, dynamic programming and – more recently – genetic algorithms. The latter type of model is particularly interesting since it mimics the evolutionary principle of selecting the “fittest” solutions from a population of candidate solutions, and then progressively evolving towards a near-optimal solution. A key feature of genetic algorithms in the road maintenance context is that some formulations allow the optimization problem to be approached at the segment and network level at the same time. Genetic algorithms are also capable of handling complex constraint sets that may not be feasible in more traditional optimization techniques. In this presentation, a highly visual and interesting introduction to genetic algorithms will be presented. The background and key elements of genetic algorithms will be presented together with some example problems and solutions. The strengths and weaknesses of genetic algorithms will be discussed in the context of road maintenance optimization. The presentation concludes with specific suggestions for how genetic algorithms can potentially be used in deterioration models in New Zealand.

Fritz Jooste has been working in the field of pavement design and asset management for more than 20 years. He earned his PhD from Texas A&M University in 1997 and since then has been involved primarily in research and with the development of systems related to pavement design and asset management. Fritz is a founder and director of Juno Services Ltd and Lonrix Ltd.

print