In order to improve the robot’s exploration efficiency of the unknown environment, this paper proposes a robot autonomous exploration algorithm based on local restricted sampling and frontier priority classification, which makes the local tree “short-sighted” by restricting the sampling range of the local tree, so that the robot can focus on exploring the vicinity of the direction of travel, avoiding repeated sampling of the explored area and improving the detection efficiency. This avoids repeated sampling of the explored area, improves the detection efficiency, and avoids frequent steering of the robot. Based on the above sampling method, by prioritizing the classification of the frontiers detected by the local tree and the global tree, the priority of the frontiers detected by the local tree is increased to ensure that the robot gives priority to exploring the frontier region detected by the local tree, and then explores the frontier detected by the global tree when the local tree fails to detect the frontier, which ensures that the unknown environment in the vicinity of the robot is sufficiently explored before exploring the more distant unknown environment. Through experimental studies in different typical indoor scenes, the results show that the algorithm proposed in this paper can effectively avoid the backtracking problem of the robot, shorten the exploration time and path, and improve the exploration efficiency, which verify the effectiveness of the algorithm.