This work introduces BioVLN, a simulation platform for developing and evaluating visual-language navigation agents in biomedical laboratories and shows that geometric exploration reaches 74.4--87.5% success, while sampling multiple valid positions in the operation area improves success and reduces unsafe proximity.
Abstract
Biomedical laboratory robots must navigate to instruments before performing experimental procedures. Existing embodied navigation platforms are designed for household environments and treat a target as an object center or an arbitrary nearby position. This representation is inadequate for laboratory instruments, which must be approached from their operating side while maintaining safe clearance from surrounding equipment. We introduce BioVLN, a simulation platform for developing and evaluating visual-language navigation agents in biomedical laboratories. BioVLN represents each instrument with three regions: its physical body, a surrounding clearance region, and an operation area in front of the usable side. This model is applied consistently to scene generation, target placement, navigation evaluation, and safety analysis, so success depends on reaching a position from which the instrument can be accessed. BioVLN supports procedural scene generation and manually designed environments, producing 47 scenes and 1667 episodes. Standardized navigation and reinforcement-learning interfaces enable trajectory collection and policy training. Experiments show that geometric exploration reaches 74.4--87.5% success, while sampling multiple valid positions in the operation area improves success to 83.3--92.5% and reduces unsafe proximity.
Vision-Language Navigation (VLN) for humanoid robots poses challenges existing benchmarks fail to address: bipedal locomotion imposes physical constraints absent from wheeled agents, humanoid morphologies vary across platforms, and egocentric observations are distorted by locomotion-induced camera dynamics. We present HumanoidVLN, a physics-grounded simulator and benchmark for VLN across diverse humanoid embodiments. Built on NVIDIA Isaac Sim, our platform supports an extensible set of humanoid configurations, demonstrated on four robots (Unitree G1, Unitree H1, Internal-A, Internal-B) spanning 10-12 lower-body DoF and heights from 1.17m to 1.80m, via a hierarchical control stack combining a reinforcement learning locomotion policy with interchangeable PD or MPC path trackers. New robots and VLN models integrate with minimal effort; we demonstrate compatibility with NaVILA, DualVLN, StreamVLN, and JanusVLN. Environments are drawn from artist-designed scenes and 3D Gaussian Splatting reconstructions, filtered for navigable areas exceeding 100 square meters. Instructions are generated by a dual generator-reviewer plus paraphraser multi-agent pipeline with human-in-the-loop verification, yielding 933 collision-aware reference episodes, each paired with one fine-grained instruction and three coarse-grained stylistic variants (formal, natural, casual). Across four models and four embodiments, JanusVLN achieves the highest mean success rate of 43.55% and nDTW of 48.38. In a 20-episode sim-to-real pilot with DualVLN and the Unitree G1, navigation errors correlate strongly (r=0.935), with a mean absolute difference of 0.68m and mean trajectory similarity of 0.782 (+/-0.188) nDTW. These results highlight the interaction between VLN models, controllers, and humanoid embodiments under physical execution. Code, benchmark, and data will be released upon acceptance at https://humanoid-vln.github.io/.
In environments where human presence is restricted due to safety concerns, cognitive robots can play a vital role in executing tasks. While robotic systems have made impressive advances in tasks such as object recognition, they still fall short in terms of true spatial understanding. Robots are not yet able to understand the 3D spatial semantics and contextual meaning of their 3D surroundings for navigation, of objects to handle in complex tasks. This research tackles this obstacle by developing a computational agent capable of learning cognitive maps from spatial data inputs in a simulated environment, mimicking the functionalities of grid and place neurons. To emulate the grid neuron's ability to generate periodic hexagonal grid-like patterns from body movements in 3-dimensional space, a novel Octant model is introduced. Additionally, a place-grid neuron interaction system is proposed to predict environmental sensations from body movements, facilitating cognitive map formation and learning mechanisms. The model was experimentally tested through a set of simulation-based experiments including: (i) a grid-based spatial arena intended to measure the accuracy of memory retrieval and encoding, (ii) a morphologically perturbed environment to measure resiliency to deformations in the trajectories, and (iii) a collection of more (object) recognition activities, to test resiliency to distortions in the object identification task. The quantitative data indicate that there is a steady high recall confidence and cosine similarity measure during the navigational trials, and a controlled growth of the place-neuron memory capacity and a strong stabilization of spatial representations in different environments. In the conditions of obstacles, the model achieves a mean Absolute Trajectory Error (ATE) of 0.42m and a root-mean-square error (RMSE) of 0.46m, which confirms that the localisation error is limited even without an explicit mapping structure. These results provide a substantive confirmation of the fact that the interference-based grid place representation allows reliable spatial localisation, mnemonic recall and navigational performance in both limited and irregular operational conditions.
A. Jain, N. Chauhan, Rahul Shrivastava· International Journal of Int...· 0 citations
We study the problem of navigating cluttered indoor environments with a humanoid robot. Unlike conventional methods that model navigation as a 2D path planning problem, humanoid traversal in cluttered environments requires continuous geometry-aware whole-body adaptation, including coordinated arm placement, torso adjustment, and gait modulation for collision-free movement through complex 3D spaces. We introduce TANGO, the first whole-body vision-language navigation framework for language-conditioned humanoid traversal in cluttered environments. Given a natural-language instruction and egocentric RGB observations, TANGO directly predicts 29-DoF joint-space actions for downstream whole-body control. We train TANGO entirely in simulation by synthesizing diverse collision-free traversal behaviors via global path planning, kinematic whole-body motion generation, obstacle-aware motion editing, and RL-based tracking. This pipeline provides dynamically feasible action supervision for learning language-conditioned whole-body policies. In extensive simulation experiments, TANGO demonstrates state-of-the-art performance in vision-language navigation, while outperforming strong modular baselines in navigating challenging scenes requiring obstacle negotiation. Lastly, we deploy TANGO zero-shot on a Unitree G1 humanoid robot, and observe robust language-guided traversal in cluttered real-world scenes without training on any real-world navigation data.
Anqi Li, Yuxin Chen, Zhaobo Li et al.· 0 citations
Autonomous mobile robots require reliable coordination among navigation, perception, tracking, and precision approach modules to complete indoor object-search missions. Existing systems often remain fragmented, treating navigation, detection, tracking, and docking as separate tasks rather than as an end-to-end pipeline. This study has two objectives: to validate a fault-tolerant coordination architecture for autonomous search-and-approach behavior and to compare two search strategies under controlled indoor target-position scenarios. The primary contribution is a methodological integration framework based on a Finite State Machine (FSM) that coordinates ROS 2 Navigation2 global navigation, YOLO11n object detection, centroid tracking, and Image-Based Visual Servoing (IBVS), while managing transitions among navigation, visual servoing, recovery, and mission-completion states. A quantitative Gazebo simulation experiment used 40 controlled trials to compare Random Exploration and Waypoint-Based Search. The integrated system achieved a 100% mission success rate without command conflicts, indicating effective FSM-based coordination between global navigation and local visual control. Waypoint-Based Search was more efficient when the target was aligned with predefined nodes, achieving a mean detection time of 50.55 s compared with 164.80 s for Random Exploration. Conversely, Random Exploration performed better when the target was away from predefined paths, reducing mean detection time to 87.00 s compared with 183.64 s. Fault-tolerant behavior was demonstrated in simulation through successful mission completion despite repeated LiDAR-triggered obstacle-recovery events during visual approach. These findings show that search efficiency depends on alignment between exploration design and spatial structure, not universal strategy superiority.
Robot foundation models (RFMs), including vision-language-action (VLA) policies, are often discussed through a scaling view: more data, larger models, and broader benchmarks should improve generalization. In robotics, however, a model can generalize while work still remains before it can run on a robot with a particular body. The work required differs across methods and target robots, and those differences affect practical deployment. We call the gap between reusable models, representations, or data and their use in execution on the target robot the embodiment gap. This survey examines what can be reused across robot embodiments and what must still be implemented on a new robot. We place existing methods on a two-axis map that shows the type of shared structure and the stage at which adaptation is needed for execution on the target robot. We then examine recent work through three overlapping research directions: sharing semantics and perception, sharing robot data and interfaces, and learning correspondence across embodiments. We also propose a reporting framework for adaptation work that success rate alone does not reveal. The framework identifies the work that should be checked when comparing cross-embodiment learning and highlights work that remains on a new robot and questions for future study.
Y. Domae, Keisuke Shirai, Hanbit Oh et al.· Trans. Mach. Learn. Res.· 0 citations
Robots deployed in delivery, campus, and emergency-response settings often need to navigate from buildings to streets within a single continuous episode. Existing benchmarks usually evaluate indoor and outdoor navigation separately, and many abstract away robot execution, leaving exit finding, boundary traversal, adaptation, and kinodynamic failures underexplored. We introduce NavVerse, a physics-enabled benchmark for indoor-to-outdoor embodied navigation. NavVerse contains 100 indoor scenes, 50 urban outdoor scenes, and 50 indoor-to-outdoor scenes, and 10,000 episodes spanning Object Navigation, Vision-and-Language Navigation, and Place Navigation tasks, where agents search for semantic points of interest such as restaurants or banks. Agents are evaluated through executable robot interfaces using task-success, path-efficiency, and safety metrics. Zero-shot experiments with RL, VLA, and modular baselines show that current agents remain far from solving cross-context navigation: end-to-end VLAs obtain the highest zero-shot success, while the modular method provides the strongest safety profile. PlaceNav further reveals a clear drop from outdoor to indoor-to-outdoor scenes, indicating that adaptation remains major bottleneck.
Junzhe Wu, Yue Hu, Zeyu Han et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
