Resumen
After private cars, minibus taxis are the most common transport mode in South Africa. Especially for low-income citizens living in townships, minibus services are often the only possibility for mobility. Despite the great importance of the mode, there is very little knowledge of routes, fares, and the number of minibuses. Hence, it is difficult to simulate and to understand the influence of this mode on other modes and on transport planning in general. This article presents the development of the first ``close-to-reality'' minibus supply model based on demand and street network only. The approach adopts the survival-of-the-fittest principle, using a co-evolutionary algorithm that is integrated into a microscopic multi-agent simulation framework. The successful application of the approach to a large-scale, real-world scenario in the Nelson Mandela Bay Area Municipality in South Africa shows that it is able to identify the main minibus corridors as well as to find robust service coverage in lower-demand areas. The resulting minibus supply model can then be used for planning purposes (e.g., to investigate aspects of strategic, operational, or regulatory changes).