Selection of road project portfolios

We describe a real-world problem from the road sector in Norway. Every four years the Norwegian Public Roads Administration (NPRA) have to select which road development projects to do the next four years. This portfolio of projects has to be selected from a much larger set of worthy projects. Todays practice in evaluating individual road projects is based on the well known cost-benefit analysis (CBA), which in our case measures the increased benefit for the society if the project is implemented. This benefit is measured in Norwegian kroner. There are many components that enter into the CBA, and not all of them are easily or naturally converted into kroner. Examples are reduction in noise level or geographic spread of the selected projects, and the effect of extra focus on these aspect should be emphasized for the decision maker. This as a multi-criteria optimization problem, where some of criteria are naturally represented as money or other scalars, and others are more naturally on an ordinal scale, using rank-based ordering. We use a tabu-search based multicriteria metaheuristic to solve this problem. This heuristic will generate a plethoria of solutions along the efficient frontier. We therefore incorporate novel ways to calculate the distance (in the solution space) between solutions, so that the user (decision maker) only is presented with sufficiently different portfolios to choose between. We will use real data from the project data base at NPRA, and compare our methods with todays practice.