In evolutionary algorithms, agents’ genotypes are often generated by more or less random mutation, followed by selection based on the fitness of their phenotypes. This paper shows that elements of this principle can be applied in multi-agent transport simulations, in the sense that a router, when faced with complex interactions between heterogeneous toll levels and heterogeneous values of time, can resort to some amount of randomness rather than being able to compute the exact best solution in every situation. The computational illustrations are based on a real world case study in the province of Gauteng, South Africa.
