Track layout design using modified genetic algorithm

Track or road layout in a given geographical area to plan new or improve existing public and/or private transportation systems is a complex problem. Especially today, many parameters have to be considered, and some of them are not obvious. For example, in face of an imminent energy crisis, the energy consumption of a transportation system should be minimized. In general, the resulting layout has to be an adequate compromise of many parameters. This means that there are many possible ways to solve the problem, which quality differs. Consequently, this type of problem implies an explosion of search-space states with raising number of midpoints and/or resolution. In these cases, heuristic search methods have their major advances compared to non-heuristic search methods, which would need time and memory proportional to the search-space size to find the best solution. We explored, in particular, the use of a genetic algorithm, as a representative of the class of evolutionary algorithms. It searches the search-space selectively by focusing on interesting regions, constantly trying to find even better regions while having significantly less memory requirements. The importance of its main parameters, their impact on the performance and the precision of the genetic algorithm are presented in this paper. Guidelines for a good set-up of the genetic algorithm are given in order to gain high efficiency and effectiveness; the encoding of parameters, the build-up of the fitness function, evaluation and reproduction of chromosomes, elitism and convergence affinity as the main engine of genetic algorithms is discussed in detail as are the limits of this approach. The modified and improved simple genetic algorithm to solve the mentioned track-layout problem forms another core part of the paper.