Path integration is one type of navigations in which navigators integrate self-motion information to update their current position and orientation relative to the origin of their travel. Human path integration ability is often measured in the path completion task. In this task, participants travel along several segments, make several turns at the intersections of each two segments, and arrive at the end of the outbound path. Then they are asked to directly return to the origin of the outbound path. Previous studies have revealed that athletes showed better path completion performance than general population. The purpose of the present study was to examine whether the path integration ability of general population can be improved if they are repeatedly exposed to outbound paths with the same configurations. In two experiments, we used the Head-Mounted Display Virtual Reality to present hallway mazes, and each outbound path consisted of 5 segments. Participants pressed a button on the gamepad to travel along a segment, so the information about transition was based on optical flow. By contrast, they were asked to actually rotate their bodies at the intersections, so the information about rotation came from both optical flow and body senses. Each participant completed 4 blocks, 6 trials of each. Within each block, they performed the path completion task on 6 different outbound paths. From one block to the next, they performed the path completion task on outbound paths with the same configurations. In Experiment 1, all the 5 segments within each outbound path had the same lengths, and the turning angle at each interaction was always 60 degree, clockwise or counterclockwise. When the participants repeatedly performed the path completion task on these outbound paths with the same configurations, they showed reduced position errors, direction errors, and RTs. By contrast, more complicated path configurations were used in Experiment 2. Specifically, within each outbound path of Experiment 2, each segment was 3 m or 5 m long, and the turning angle at each intersection was 60 or 120 degree, clockwise or counterclockwise. The participants also showed recued RTs and percentage position errors when these complex configurations were repeated. What is more, the absence of the sex difference in both experiments revealed that both males and females can benefit from the exposure to the repeated configurations. Taken together, these results demonstrated that the path integration ability of general population can be improved if they have repeatedly performed the path completion task on the outbound paths with the same configurations, though more repetitions are needed when the path configurations are more complex. These results revealed the influence of previous experience on human navigation, and such influence of spatial configuration on human path integration might be similar to that of implicit learning in visual cognition. Future research is needed to further investigate the training of human path integration ability.