Generating sequences of actions (trajectories) that an agent takes to solve a task, used for training via imitation learning.