Batch and semi-batch processes provide a flexible means of producing high value-added products in the chemical, biotechnical, and pharmaceutical industries. Unlike continuous processes, batch processes are inherently transient and typically also nonlinear, and a nonlinear dynamic end-point optimization problem needs to be solved in order to determine the optimal operating strategy. Inter-and intra-run variations and lack of better measurement information typically only allow us to build an imperfect model of the process, and the remaining uncertainties can be large. Nominal optimization techniques, requiring an exact process model, may thus not always be suitable for batch process optimization. This contribution suggests an alternative approach in which modeling and identification uncertainty are explicitly accounted for during the process optimization. As opposed to a deterministic quantity that is a function of only the nominal model, a probabilistic measure of success is optimized, leading to robustness of the desired objective to uncertainties and variations. Both simulation and experimental results are given to demonstrate the idea and the application of the proposed approach and to highlight the benefits that can be expected from its industrial application.