A method for generating a training data set for training an artificial intelligence (AI) module. An image sequence is provided in which surroundings of a robot are recorded. A trajectory in the recorded surroundings is determined. At least one future image sequence is generated which extends to a time segment in the future, and, based on the at least one determined trajectory, encompasses a prediction of images for the event that the determined trajectory was followed during the time segment in the future. At least one sub-section of the determined trajectory in the generated image sequence is assessed as positive or as negative when a movement predicted by following the trajectory corresponds to a valid movement situation, or as an invalid movement situation, respectively. The generated future image sequence with the assessment assigned thereto of the trajectory are combined for generating a training data set for the AI module.