Mapping is one of the crucial tasks enabling autonomous navigation of a mobile robot. Conventional mapping methods output a dense geometric map representation, e.g. an occupancy grid, which is not trivial to keep consistent for prolonged runs covering large environments. Meanwhile, capturing the topological structure of the workspace enables fast path planning, is typically less prone to odometry error accumulation, and does not consume much memory. Following this idea, this letter introduces PRISM-TopoMap – a topological mapping method that maintains a graph of locally aligned locations not relying on global metric coordinates. The proposed method involves original learnable multimodal place recognition paired with the scan matching pipeline for localization and loop closure in the graph of locations. The latter is updated online, and the robot is localized in a proper node at each time step. We conduct a broad experimental evaluation of the suggested approach in a range of photo-realistic environments and on a real robot, and compare it to state of the art. The results of the empirical evaluation confirm that PRISM-Topomap consistently outperforms competitors computationally-wise, achieves high mapping quality and performs well on a real robot.

In recent years, Embodied AI has become one of the main topics in robotics. For the agent to operate in human-centric environments, it needs the ability to explore previously unseen areas and to navigate to objects that humans want the agent to interact with. This task, which can be formulated as ObjectGoal Navigation (ObjectNav), is the main focus of this work. To solve this challenging problem, we suggest a hybrid framework consisting of both not-learnable and learnable modules and a switcher between them—SkillFusion. The former are more accurate, while the latter are more robust to sensors’ noise. To mitigate the sim-to-real gap, which often arises with learnable methods, we suggest training them in such a way that they are less environment-dependent. As a result, our method showed top results in both the Habitat simulator and during the evaluations on a real robot.

Mapping is one of the key components of mobile robot navigation. Representing a map as a topological structure is suitable for fast path planning and does not require high positioning precision or high computational resources, which is particularly useful in large environments. In recent years, numerous methods of topological graph building have emerged. Most of these methods build a topological graph in conjunction with a metric map, and during tests and benchmarks, the performance of metric maps is typically assessed. However, topological graph quality is also crucial for robot navigation. In this study, we conducted an extensive evaluation of five open-source topological mapping methods (Hydra, S-graphs+, IncrementalTopo, ETPNav, and TSGM) using a large simulated indoor dataset with precise and noisy positioning. We used novel metrics to measure topological graph quality in our comparison. We found that all the methods except TGSM build an unconnected graph with both precise and noisy positioning. Hydra and S-graphs are more robust to position noise, but their graphs have a high percent of the edges crossing an obstacle. TSGM builds connected graphs, and it is not sensitive to positional noise because it does not use position as an input. However, its graphs have a high percent of the edges crossing an obstacle. IncrementalTopo has fewer edges passing through the obstacles; however, with noisy positioning, its graph becomes highly disconnected, which may also cause navigation failures. Thus, all the evaluated methods have their own advantages and drawbacks, and none of them builds a graph suitable for navigation out of the box. Overall, our study pinpoints the advantages and disadvantages of the state-of-the-art topological mapping methods and may aid researchers and practitioners in choosing a proper available algorithm for their specific setup.