Train trajectory optimization for energy-efficient and minimum-time train control has attracted the attention of many researchers and practitioners, due to its potential in driver advisory systems and in automatic train operation. For algorithmic developments in the field to have an impact on the railway industry, it is evident that low computation times are of paramount importance. Various methods based on Pontryagin’s maximum principle and dynamic programming have been proposed in the literature that satisfy this requirement, each with their advantages and limitations. In this paper, we use state-of-the-art methods and software from the field of embedded optimization to repeatedly solve trajectory optimization problems as the train progresses on the track. We demonstrate how challenging problem formulations can be tackled in milliseconds, even on resource-constrained hardware, and study the effect of feedback delay on the control performance.