Jul 2026· Italian National Conference on Sensors· Vol 26, pp. 4674· 0 citations· 27 references
Medicine
TL;DR
The proposed PC-IMVL integrates deep visual perception with embedded physical modeling, which explicitly introduces spatial physical constraints into the network optimization process and builds a physical consistency-aware loss function to regularize 3D position and pose estimation, which enables efficient and reliable embedded deployment.
Abstract
Accurate object localization is essential for enabling autonomous operation of indoor robotic systems. As a low-cost, compact, and flexibly deployable solution, monocular visual object localization (MVOL) is highly applicable to lightweight embedded robotic platforms. However, conventional data-driven MVOL methods suffer from inherent limitations of 2D-to-3D ill-posed mapping, which is caused by the incapability of constraining the spatial physical logic of real scenes, resulting in severe depth ambiguity, inaccurate scale estimation, and physically unreasonable predictions. To address these issues, this paper proposes a novel physics-guided monocular visual localization framework termed PC-IMVL for indoor scenarios. The PC-IMVL integrates deep visual perception with embedded physical modeling, which explicitly introduces spatial physical constraints into the network optimization process and builds a physical consistency-aware loss function to regularize 3D position and pose estimation. Combined with a lightweight tailored architecture, the framework enables efficient and reliable embedded deployment. Offline experiments and real-world online tests validate the effectiveness of the proposed method. PC-IMVL yields average absolute errors (AE) of 0.095–0.333 m, reducing the localization error of early fusion methods by more than 50%. Within a working distance of 3–4 m, it achieves a relative error (RE) of 2.4% and a horizontal viewing angle error (VAE) below 2°, outperforming existing state-of-the-art MVOL methods. The effectiveness of the physical guidance mechanism is verified. This work provides a practical high-precision localization solution for embedded indoor robotic systems.
Cameras are ubiquitous sensors in robotics due to their compact form factor and the perceptual richness captured through visual information. Monocular SLAM enables robots to understand the environment with a minimum setup, however, it inherently suffers from scale ambiguity. A common solution is to provide multi-modal sensor configurations, such as visual-inertial systems, where scale is observable unless the robot navigates under a constant-velocity motion, a common scenario in mobile robotics. With the advent of deep-learning, geometric foundation models have been used to address this problem, but the depths maps are often noisy and scale-inconsistent across frames. In this paper, we propose Scalix, a real-time monocular SLAM framework that achieves metric-scale state estimation by integrating learned depth cues into a probabilistic factor-graph formulation. By augmenting existing monocular depth models with both per-pixel depth uncertainty and per-frame scale uncertainty, Scalix treats scale predictions as independent measurements within its optimization, leading to improved scale consistency through multi-view data associations. Experiments in large-scale outdoor and indoor environments demonstrate state-of-the-art performance on both metric and up-to-scale benchmarks while maintaining real-time operation and generalization.
: Visual simultaneous localization and mapping (VSLAM) is a key technology for mobile robotics, autonomous driving, and embodied intelligence, enabling self-localization, environment reconstruction, and scene understanding. Although conventional geometric methods have achieved notable success, their performance often degrades in challenging conditions, such as low-texture scenes, severe illumination changes, dynamic interference, and long-term environmental variations. Recent advances in deep learning have created new opportunities to improve VSLAM through stronger feature representations, learned priors, semantic perception, and emerging map representations. At the same time, the increasing adoption of learning-based modules has raised important questions about integration strategies, generalization, interpretability, and real-time deployment. This paper presents a systematic review of deep learning-enhanced VSLAM, with a particular focus on how learning models are incorporated into classical simultaneous localization and mapping (SLAM) pipelines and how they function within the overall system. To provide a unified perspective, existing methods are organized into five categories according to their fusion interfaces with geometric SLAM pipelines: observation-level interfaces, constraint/prior/weight-level interfaces, solver-level interfaces, representation-level interfaces, and system-level integration interfaces. Based on this taxonomy, representative approaches are comparatively analyzed for accuracy, robustness, efficiency, and deployability. In addition, this review summarizes common design principles, including geometric consistency constraints, error propagation characteristics, and typical failure modes, and further discusses open challenges and future directions such as lightweight deployment, cross-domain adaptation, dynamic map modeling, and long-term consistency maintenance. This review aims to provide a structured reference for the analysis, design, and deployment of learning-enhanced VSLAM systems.
Xiruo Chen, Qi Ouyang, Sihong Meng et al.· Computers, Materials & C...· 0 citations
Monocular depth estimation has achieved strong open-domain generalization, yet reliable robotic deployment remains difficult in transparent, reflective, and specular environments, where depth sensors often produce missing or biased depth. Existing methods often handle such optical failures with scene-specific preprocessing, auxiliary modules, or post-hoc fine-tuning. While effective in constrained settings, these designs increase architectural redundancy and can over-specialize general geometry models to narrow optical scenarios. We revisit this problem as a localized failure mode within base-model training and identify sensor-induced supervision bias as a key bottleneck: models inherit sensor failure patterns from biased real-depth supervision in optically challenging regions. We then introduce OptiGeo, a bias-aware training framework that rehabilitates biased real supervision using a clean-geometry teacher and residual-trimmed alignment. We redefine transparency-targeted rendering as a compact source of clean optical geometry, rather than a large domain-specific fine-tuning set. With only a small targeted rendering set, OptiGeo learns the geometric structure of transparent objects and regions, correcting local geometry distortions that real sensors cannot reliably supervise. Despite only 30M parameters, OptiGeo outperforms substantially larger 300M-scale monocular models and billion-scale multi-view baselines on transparent-scene benchmarks, while remaining competitive on general zero-shot depth and boundary sharpness. Real-world navigation cases further validate its practicality as an efficient perception module in optically challenging scenes.
Mu-Xin Liu, Tian-Bo Liu, Jing Xia et al.· 1 citation
Large-scale indoor mapping and positioning with vision sensors is fundamental to a wide range of applications, such as robotic navigation and augmented reality. However, the rapidly increasing number of detectable objects and the expanded spatial coverage jointly introduce matching ambiguity and high computational cost. Fine-grained object maps can improve accuracy but often accumulate redundant observations and slow down localization, whereas overly compressed scene representations may discard essential semantic and structural cues and degrade robustness. To balance accuracy and efficiency for indoor spatial sensing, we propose TS-MapLoc, a map-centric object-level localization framework based on cross-layer semantic co-mapping. It builds a lightweight topological–semantic map that integrates multi-scale information from the image layer and the object layer, reducing redundancy while preserving key structural constraints. On top of this map, a cognition-inspired progressive localization strategy performs coarse-to-fine inference via stage-wise filtering under cross-layer semantic consistency, effectively narrowing the search space and stabilizing matching. The proposed method supports efficient and accurate object-level localization for built-environment applications.
Cui-Yun Fang, Fan Wang, Ye-Dong Jiang et al.· IEEE Signal Processing Lette...· 0 citations
Self-supervised monocular depth estimation has emerged as an appealing solution to design lightweight and effective models for deployment on computationally constrained devices due to its reduced reliance on expensive depth sensors. By eliminating the need for ground-truth annotations and leveraging the simplicity of monocular camera setups, this approach facilitates cost-effective data collection and broad applicability across fields such as computer vision and robotics. A critical challenge is achieving resource-efficient neural networks without compromising the overall performance. State-of-the-art models generally adopt depth-wise convolutions and attention mechanisms; however, these functions often incur high energy costs and face compatibility issues in embedded environments. To address this, we propose XiDepth, a lightweight architecture based on the XiNet operator block, designed to enhance feature extraction while maintaining low computational complexity and energy demand. On the KITTI dataset, XiDepth achieves state-of-the-art performance with only 0.8M parameters. Tests on a Raspberry Pi 4 further confirm its suitability for real-world embedded applications, reducing FLOPs by 40% and energy consumption by 35% compared to leading methods.
Elena Izzo, Riccardo Toniolo, Lamberto Ballan· 1 citation
Localization is a fundamental requirement for autonomous mobile robot navigation. Several localization techniques exist, but they often require extensive installation of beacons, careful parameter tuning, high computational requirements, or an immense amount of training data. Environmental (e.g., indoor)/resource constraints, sensor degradation, and sudden pose discontinuities can make such methods unreliable. This creates a critical gap: the lack of a simple, lightweight localization method that can operate as a primary localization method or in parallel with other classical systems and provide reliable pose estimates during primary localization system failures. Thus, this paper proposes a lightweight, deep learning (DL)-based, two-dimensional LiDAR localization method. The approach combines LiDAR scan range data with eleven proposed handcrafted geometric features to train a Convolutional Multi-Layer Perceptron (ConvMLP) regression model for predicting the two-dimensional location of a robot, which is further smoothed by an augmented recursive Extended Kalman filter (EKF). The overall system is validated in three real-world environments. The results are compared against various existing machine learning (ML) models and other well-known localization techniques. The experimental results demonstrate a 280 Hz pose-update rate, achieving a 13 cm Root Mean Square Error (RMSE) using the ConvMLP model alone, which further reduces to 5 cm when fused with the recursive EKF.
Harsha Keladi Ganapathi, S. Mukhopadhyay· Applied Sciences· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
