Sub-trajectory clustering method and device based on reinforcement learning

One or more embodiments of the invention provide a reinforcement learning-based sub-trajectory clustering method and device. The method comprises the steps that a to-be-divided trajectory set is acquired, the trajectory set comprises at least one trajectory, and the at least one trajectory comprises k trajectories used as initial clustering centers; performing reinforcement learning on the at least one track, traversing each track point included in each track so as to decide a division action corresponding to each track point, and dividing the track where each track point is located by adopting the division action corresponding to each track point so as to obtain at least one sub-track, clustering is carried out on the sub-tracks obtained through division to obtain k sub-track sets, so that values of clustering evaluation indexes of the k sub-track sets obtained through clustering conform to expected values; and obtaining k sub-track sets obtained based on reinforcement learning, and taking the k sub-track se