T2Nav Algebraic Topology Aware Temporal Graph Memory and Loop Detection for ZeroShot Visual Navigation
Authors: Quang-Anh N. D., Duc Pham, Minh-Anh Nguyen, Tung Doan, Tuan Dang
Organizations: International School, Vietnam National University, Hanoi, Vietnam. · *Cognitive Robotics Lab, Department of Electrical Engineering and Computer Science, University of Arkansas, Fayetteville, AR, USA. · Hanoi University of Science, Vietnam National University, Hanoi, Vietnam. · Hanoi University of Science and Technology, Hanoi, Vietnam.
Deploying autonomous agents in real world environments is challenging, particularly for navigation, where systems must adapt to situations they have not encountered before. Traditional learning approaches require substantial amounts of data, constant tuning, and, sometimes, starting over for each new task. That makes them hard to scale and not very flexible. Recent breakthroughs in foundation models, such as large language models and vision language models, enable systems to attempt new navigation tasks without requiring additional training. However, many of these methods only work with specific input types, employ relatively basic reasoning, and fail to fully exploit the details they observe or the structure of the spaces. Here, we introduce T2Nav, a zeroshot navigation system that integrates heterogeneous data and employs graph-based reasoning. By directly incorporating visual information into the graph and matching it to the environment, our approach enables the system to strike a good balance between exploration and goal attainment. This strategy allows robust obstacle avoidance, reliable loop closure detection, and efficient path planning while eliminating redundant exploration patterns. The system demonstrates flexibility by handling goals specified using reference images of target object instances, making it particularly suitable for scenarios in which agents must navigate to visually similar yet spatially distinct instances. Experiments demonstrate that our approach is efficient and adapts well to unknown environments, moving toward practical zero-shot instance-image navigation capabilities.
Figures & tables
Fig. 1: Conceptual overview of T 2 -Nav. The framework addresses loop-closure problems in which the agent initially becomes trapped in repetitive exploration patterns. To address this issue, we propose two novel modules: (1) TeRM maintains cross-temporal object relationships, and (2) scene dynamics are explored through graph-based reasoning for TSLC to employ persistent homology, enabling the detection and avoidance of navigation loops via topological invariants of agent trajectories, thereby facilitating a robust zero-shot visual navigation system.
Fig. 2: Overview of T2 -Nav system. (a) Multi-modal inputs with GPS, pose, and goal image specifying the target instance, and RGBD input to construct a dynamic scene graph. (b) Graph Processing with RGBD images performs scene-goal matching to identify potential target instances, with ✓ / × indicating success/failure. (c) Loop closure detection maintains a blacklist of locations. (d) TeRM provides temporal consistency for robust instance tracking. (d) The TeRM module maintains temporal consistency across consecutive scene graphs for robust instance tracking across varying viewpoints and environmental conditions. (e) The navigation pipeline generates occupancy maps, applies a deterministic local policy for obstacle avoidance, and outputs action decisions.
Fig. 3: TeRM Framework Overview. (a) Scene graphs track object persistence and new entities across timesteps. (b) Temporal edge construction for dependencies between nodes using exponentially decayed edge weights across t . (c) Physics-inspired positions prediction refines object trajectories based on motion dynamics.
Fig. 4: TSLC illustration: Agent trajectories are processed through Vietoris-Rips filtration to derive persistent homological features, which are subsequently mapped to a persistence diagram in a birth-death coordinate system. Loop closure is identified by assessing whether the W2 distance between the persistence diagram of the current trajectory segment and those of historical segments falls below a predefined threshold.
Method
Training-Free
Input
SR ↑
SPL ↑
Krantz et al. [ 2 ]
×
RGBD ++
8.3
3.5
ZSON [ 26 ]
×
RGB +
14.6
7.3
OVRL-v2 [ 27 ]
×
RGB ++
24.8
11.8
IEVE [ 28 ]
×
RGBD ++
70.2
25.2
PSL [ 29 ]
×
RGB -
23.0
11.4
GOAT [ 30 ]
×
RGBD
37.4
16.1
TABLE I: Instance-image-goal navigation results on HM3D, methods with training-free capability are marked with ✓ , where - , + , and ++ denote only orientation input, only coordinate input, and for both of them, respectively.
Fig. 5: Qualitative comparison of frontier selection strategies between UniGoal and the proposed method. Compared with the baseline UniGoal, T2 -Nav selects frontiers more strategically, for more efficient exploration and reducing redundant path traversals.
Fig. 6: Qualitative comparison of navigation trajectories between UniGoal and the proposed method. Compared with the baseline UniGoal, T 2 -Nav produces shorter, more efficient trajectories with fewer detours, thereby demonstrating improved goal-directed navigation.
Method
SR ↑
SPL ↑
Baseline [ 17 ]
66.16
24.83
w/o Temporal Reasoning Memory Module
74.99
28.28
w/o Topological Signatures for Loop Closure
72.22
25.48
Proposed T2 -Nav
75.62
28.38
TABLE II: Ablation results on IIN over 200 episodes on HM3D.
Navigating to instance-level targets in complex environments is a challenging problem. Many existing zero-shot methods achieve strong performance by modeling the entire environment and leveraging large language models for scene understanding. However, such strategies primarily focus on exploring new regions while lacking a deeper exploitation of information from previously explored areas. Consequently, when targets are missed or misidentified within previously visited regions, navigation failures occur frequently. To address these limitations, we propose MCNav, a memory-aware navigation framework with a dynamic cognitive map. This map stores efficiently queryable information about relevant objects in explored areas. Building on this memory structure, MCNav introduces two memory-aware exploration strategies: goal re-validation, which re-assesses previously seen objects to correct matching failures, and missed goal re-exploration, which estimates the likelihood that a target is present in an explored region from contextual cues. These strategies are further stabilized by a blacklist mechanism to prevent repeated errors and a double-check mechanism for high-confidence confirmation. We evaluate MCNav on the HM3Dv1 and HM3Dv2 datasets across three different tasks, where it achieves state-of-the-art performance, particularly on the instance-level goal navigation task.
Jingyu Li, Zhe Liu, Wenxiao Wu +1
Fudan University · Shanghai Innovation Institute · University of Hong Kong +1
Zero-shot object navigation requires agents to locate unseen targets in unfamiliar environments without prior maps or task-specific training. Despite the commonsense reasoning ability of vision-language models (VLMs), existing mapless navigators often suffer from repeated exploration and premature stopping caused by limited historical context and false-positive target predictions. We propose ReMemNav, a training-free framework that combines lightweight semantic grounding, memory-based decision correction, and target verification. RAM-derived semantic priors guide panoramic direction selection, while a geometry-triggered correction mechanism retrieves historical scene descriptions when the selected direction points toward previously visited regions. When a target is predicted to be visible, ReMemNav verifies the corresponding single-view observation before the final approach. Depth-based action sampling then converts the high-level decision into collision-free motion. Experiments on HM3D and MP3D show that ReMemNav achieves higher success rates and path efficiency than existing training-free zero-shot baselines. Specifically, ReMemNav achieves absolute SR/SPL gains of 1.7%/5.3% on HM3D v0.1, 18.2%/11.1% on HM3D v0.2, and 8.7%/6.9% on MP3D.
Zero-shot object-goal navigation aims to find target objects in unseen environments using only egocentric observation. Recent methods leverage foundation models' comprehension and reasoning capabilities to enhance navigation performance. However, when faced with poor viewpoints or weak semantic cues, foundation models often fail to support reliable reasoning in both perception and planning, resulting in inefficient or failed navigation. We observe that inherent relationships among objects and regions encode structured scene priors, which help agents infer plausible target locations even under partial observations. Motivated by this insight, we propose Spatial Relation-aware Navigation (SR-Nav), a framework that models both observed and experience-based spatial relationships to enhance both perception and planning. Specifically, SR-Nav first constructs a Dynamic Spatial Relationship Graph (DSRG) that encodes the target-centered spatial relationships through the foundation models and updates dynamically with real-time observations. We then introduce a Relation-aware Matching Module. It utilizes relationship matching instead of naive detection, leveraging diverse relationships in the DSRG to verify and correct errors, enhancing visual perception robustness. Finally, we design a Dynamic Relationship Planning Module to reduce the planning search space by dynamically computing the optimal paths based on the DSRG from the current position, thereby guiding planning and reducing exploration redundancy. Experiments on HM3D show that our method achieves state-of-the-art performance in both success rate and navigation efficiency. The code will be publicly available at https://github.com/Mzyw-1314/SR-Nav
Leyuan Fang, Zan Mao, Zijing Wang +1
School of Artificial Intelligence and Robotics, Hunan University, Changsha, China, 410114, China