ROVER: Robust Loop Closure Verification With Trajectory Prior in Repetitive Environments

Loop closure detection is important for simultaneous localization and mapping (SLAM), which associates current observations with historical keyframes, achieving drift correction and global relocalization. However, a falsely detected loop can be fatal, and this is especially difficult in repetitive environments where appearance-based features fail due to the high similarity. Therefore, verifying a loop closure is a critical step to avoid false-positive detections. Existing works in loop closure verification predominantly focus on learning invariant appearance features, neglecting the prior knowledge of the robot’s spatial-temporal motion cue, i.e., trajectory. In this article, we propose ROVER, a loop closure verification method that leverages the historical trajectory as a prior constraint to reject false loops in challenging repetitive environments. For each loop candidate, it is first used to estimate the robot trajectory with pose-graph optimization. This trajectory is then submitted to a scoring scheme that assesses its compliance with the trajectory without the loop, which we refer to as the trajectory prior constraint (TPC), to determine if the loop candidate should be accepted. Benchmark comparisons and real-world experiments demonstrate the effectiveness of the proposed method. Furthermore, we integrate ROVER into state-of-the-art SLAM systems to verify its robustness and efficiency. Our source code and self-collected dataset are available at https://rover-lcv.github.io/. Note to Practitioners—This paper addresses a critical reliability challenge for autonomous mobile robots operating in repetitive environments, such as large warehouses, long corridors, or symmetrical structures. In these settings, high appearance similarity often causes robots to misidentify distinct locations. If the localization system incorrectly merges these locations (a “false loop closure”), it can severely corrupt the robot’s map and trajectory, leading to navigation failures that require manual intervention. The primary contribution of this work is ROVER, a verification method that acts as a safeguard for simultaneous localization and mapping (SLAM) systems. Unlike traditional methods that rely primarily on appearance to validate loop closures, ROVER assesses the geometric consistency of the robot’s historical motion. It effectively simulates the optimization that would occur if a match were accepted; if the resulting trajectory contradicts the robot’s known motion history—implying physically impossible jumps or distortions—the match is rejected. For practitioners, this approach offers a cost-effective method to enhance system robustness without requiring additional sensors or computationally expensive semantic understanding. While the system integration presented focuses on vision-based systems, the core methodology is sensor-agnostic and compatible with existing graph-based SLAM backends. It is important to note that ROVER relies on uncertainty estimates (covariance) for the robot’s trajectory, which can be reliably estimated by practitioners for typical SLAM systems. The integration into proprietary navigation stacks should be straightforward with the provided source code to facilitate rapid prototyping and deployment.