Large-scale Outdoor Environments Research Articles

How can children best learn map skills? a step-by-step approach

ABSTRACT This study contributes to our understanding of children's development of map-reading skills through a systematic ‘step-by-step approach’. Utilising an ecological research design, we investigated how children perceive and utilise maps in a large-scale outdoor environment. A total of 156 children, aged 5–11 (both boys and girls), participated in the study, tasked with locating ten controls in the terrain using five different types of maps. The study measured accuracy in wayfinding (the number of correctly identified controls) and efficiency in wayfinding (the time spent on the task). The results revealed that children exhibited better mastery of perspective maps compared to symbolic maps, with older children demonstrating significantly greater proficiency with symbolic maps than younger children. Notably, there were no significant differences in performance between boys and girls.

EFRNet-VL: An end-to-end feature refinement network for monocular visual localization in dynamic environments

This study addresses the challenge of visual localization using monocular images, a crucial technology for autonomous systems that facilitates their navigation and interaction capabilities. With the advent of deep learning, visual localization techniques that utilize these methods have demonstrated improved robustness across diverse environments. Existing end-to-end models apply convolutional neural networks (CNNs) to extract salient features and directly estimate continuous spatial poses from map models that allow for implicit differentiation. Nonetheless, these models often falter in adapting their feature representations to extreme variations in environmental conditions, leading to critical localization inaccuracies during episodes of altered lighting, varying weather, or in the presence of moving objects. To overcome these limitations, we introduce the end-to-end feature refinement network for visual localization (EFRNet-VL). This network architecture is specifically designed to prioritize the extraction of static features crucial for the six degrees of freedom (6DoF) pose estimation, thereby outperforming prior methodologies. EFRNet-VL meticulously integrates a convolutional network structure with self-attention mechanisms and Long Short-Term Memory (LSTM) modules, which together facilitate the accurate association of a single image with its corresponding camera pose, even within dynamic environments. The proposed feature refinement approach is straightforward to implement and can enhance the performance of existing neural pose estimators. Our comprehensive evaluations of EFRNet-VL underscore its effectiveness. Notably, it has diminished the average position and orientation errors by 54.5% and 25.7%, respectively, as compared to the popular PoseNet model across various indoor settings. Moreover, in large-scale outdoor environments, it has achieved an average localization precision of 7.02m/2.79°. EFRNet-VL has set a new benchmark for end-to-end learning-based methods in visual localization and operates efficiently in real time, processing at a speed of 9.8 ms per image frame.

A Visual Positioning Method of UAV in a Large-Scale Outdoor Environment.

Visual positioning is a basic component for UAV operation. The structure-based methods are, widely applied in most literature, based on local feature matching between a query image that needs to be localized and a reference image with a known pose and feature points. However, the existing methods still struggle with the different illumination and seasonal changes. In outdoor regions, the feature points and descriptors are similar, and the number of mismatches will increase rapidly, leading to the visual positioning becoming unreliable. Moreover, with the database growing, the image retrieval and feature matching are time-consuming. Therefore, in this paper, we propose a novel hierarchical visual positioning method, which includes map construction, landmark matching and pose calculation. First, we combine brain-inspired mechanisms and landmarks to construct a cognitive map, which can make image retrieval efficient. Second, the graph neural network is utilized to learn the inner relations of the feature points. To improve matching accuracy, the network uses the semantic confidence in matching score calculations. Besides, the system can eliminate the mismatches by analyzing all the matching results in the same landmark. Finally, we calculate the pose by using a PnP solver. Furthermore, we evaluate both the matching algorithm and the visual positioning method experimentally in the simulation datasets, where the matching algorithm performs better in some scenes. The results demonstrate that the retrieval time can be shortened by three-thirds with an average positioning error of 10.8 m.

Visual–Inertial Navigation System Based on Virtual Inertial Sensors

In this paper, we propose a novel visual–inertial simultaneous localization and mapping (SLAM) method for intelligent navigation systems that aims to overcome the challenges posed by dynamic or large-scale outdoor environments. Our approach constructs a visual–inertial navigation system by utilizing virtual inertial sensor components that are mapped to the torso IMU under different gait patterns through gait classification. We apply a zero-velocity update (ZUPT) to initialize the system with the original visual–inertial information. The pose information is then iteratively updated through nonlinear least squares optimization, incorporating additional constraints from the ZUPT to improve the accuracy of the system’s positioning and mapping capabilities in degenerate environments. Finally, the corrected pose information is fed into the solution. We evaluate the performance of our proposed SLAM method in three typical environments, demonstrating its applicability and high precision across various scenarios. Our method represents a significant advancement in the field of intelligent navigation systems and offers a promising solution to the challenges posed by degenerate environments.

DCPCR: Deep Compressed Point Cloud Registration in Large-Scale Outdoor Environments

Reliable and accurate registration of point clouds is a challenging problem in robotics as well as in the domain of autonomous driving. In this article, we address the task of aligning point clouds with low overlap, containing moving objects, and without prior information about the initial guess. We enhance classical ICP-based registration with neural feature-based matching to reliably find point correspondences. Our novel 3D convolutional and attention-based network is trained in an end-to-end fashion to learn features, which are well suited for matching and for rating the quality of the point correspondences. By utilizing a compression encoder, we can directly operate on a compressed map representation, making our approach well suited for operation under memory constraints. We evaluate our approach on point clouds obtained at completely different points in time, showing that our approach is able to register point clouds even under those challenging conditions reliably.

LiDAR-Inertial-GNSS Fusion Positioning System in Urban Environment: Local Accurate Registration and Global Drift-Free

Aiming at the insufficient accuracy and accumulated error of the point cloud registration of LiDAR-inertial odometry (LIO) in an urban environment, we propose a LiDAR-inertial-GNSS fusion positioning algorithm based on voxelized accurate registration. Firstly, a voxelized point cloud downsampling method based on curvature segmentation is proposed. Rough classification is carried out by the curvature threshold, and the voxelized point cloud downsampling is performed using HashMap instead of a random sample consensus algorithm. Secondly, a point cloud registration model based on the nearest neighbors of the point and neighborhood point sets is constructed. Furthermore, an iterative termination threshold is set to reduce the probability of the local optimal solution. The registration time of a single frame point cloud is increased by an order of magnitude. Finally, we propose a LIO-GNSS fusion positioning model based on graph optimization that uses GNSS observations weighted by confidence to globally correct local drift. The experimental results show that the average root mean square error of the absolute trajectory error of our algorithm is 1.58m on average in a large-scale outdoor environment, which is approximately 83.5% higher than that of similar algorithms. It is fully proved that our algorithm can realize a more continuous and accurate position and attitude estimation and map reconstruction in urban environments.

RadarSLAM: A robust simultaneous localization and mapping system for all weather conditions

A Simultaneous Localization and Mapping (SLAM) system must be robust to support long-term mobile vehicle and robot applications. However, camera and LiDAR based SLAM systems can be fragile when facing challenging illumination or weather conditions which degrade the utility of imagery and point cloud data. Radar, whose operating electromagnetic spectrum is less affected by environmental changes, is promising although its distinct sensor model and noise characteristics bring open challenges when being exploited for SLAM. This paper studies the use of a Frequency Modulated Continuous Wave radar for SLAM in large-scale outdoor environments. We propose a full radar SLAM system, including a novel radar motion estimation algorithm that leverages radar geometry for reliable feature tracking. It also optimally compensates motion distortion and estimates pose by joint optimization. Its loop closure component is designed to be simple yet efficient for radar imagery by capturing and exploiting structural information of the surrounding environment. Extensive experiments on three public radar datasets, ranging from city streets and residential areas to countryside and highways, show competitive accuracy and reliability performance of the proposed radar SLAM system compared to the state-of-the-art LiDAR, vision and radar methods. The results show that our system is technically viable in achieving reliable SLAM in extreme weather conditions on the RADIATE Dataset, for example, heavy snow and dense fog, demonstrating the promising potential of using radar for all-weather localization and mapping.

Online Semantic-Assisted Topological Map Building With LiDAR in Large-Scale Outdoor Environments: Toward Robust Place Recognition

The building of map model is crucial to the efficiency of unmanned ground vehicles’ (UGVs) navigation and robust place recognition. This paper presents an online semantic-assisted topological map building approach using 3-D LiDAR point clouds in large-scale outdoor environments. To achieve global consistency while switching between global navigation satellite system (GNSS)-denied and GNSS-fixed scenes, a pose estimation module composed of semantic-assisted LiDAR-simultaneous localization and mapping (SLAM) and the GNSS positioning is designed and integrated seamlessly, which can supply both the UGV’s global 6-degrees of freedom (DoF) poses and the local semantic-metric map. On the basis of semantic-metric map output, an online topological map building framework is implemented aiming to construct the topological map with robust place recognition performance, in which a novel local node density adjustment (LoNDA) algorithm is proposed to online adjust the distance between adjacent nodes based on both the targeted node density and appearance similarity. Field experiments are conducted on both self-collected datasets and public datasets. Experimental results demonstrate that the proposed approach can perform the online LiDAR-based topological map building with high accuracy, and the proposed LoNDA algorithm can achieve superior place recognition performance than the density-based scheme.

A VINS Combined With Dynamic Object Detection for Autonomous Driving Vehicles

This paper proposes a visual-inertial navigation system (VINS) combined with a dynamic object detection (DOD) algorithm, to improve the localization and state estimate accuracy of autonomous driving vehicles (ADV) in dynamic environments. Firstly, based on the YOLOv5 network, we train the proposed DOD model to detect dynamic objects in the road environment. Secondly, by removing the feature points in the dynamic object regions, we track the remaining feature points to eliminate the influence of the dynamic object. Furthermore, we model the global positioning system (GPS) measurement as a general factor and introduce its residual factor into the cost function to eliminate the cumulative error. Finally, we validate the performance of the proposed method on public datasets and real-world experiments. The results show that the proposed method can effectively eliminate the influence of the dynamic object and eliminate the cumulative error. It provides theoretical guidance for ADV navigation in dynamic or large-scale outdoor environments.

Multiview vision-based human crowd localization for UAV fleet flight safety

This paper presents a centralized, vision-based method for robust, on-the-fly 3D localization and mapping of human crowds in large-scale outdoor environments, assuming their independent visual detection on the camera feed of multiple UAVs. The proposed method aims at enhancing vision-assisted human crowd avoidance, in line with common UAV safety regulations, since the resulting 3D crowd annotations may be employed by other algorithms for on-line mission/path replanning during deployment of a UAV fleet. Initially, 2D crowd heatmaps are assumed to be derived per video frame on-board each UAV separately, using deep neural human crowd detectors, which indicate the probability of each pixel depicting a human crowd. The UAV-mounted cameras are assumed to be covering the same large-scale outdoor area over time. The heatmaps of each time instance are transmitted to a central computer and back-projected onto the common 3D terrain/map of the navigation environment, utilizing the intrinsic and extrinsic camera parameters. The projected crowd heatmaps derived from the different drones/cameras are fused by exploiting a Bayesian filtering approach that favors newer crowd observations over older ones. Thus, during flight, an area is marked as crowded (therefore, a no-fly zone) if all, or most, individual UAV-mounted visual detectors have recently and confidently indicated crowd existence on it. In order to calculate prior probabilities for Bayesian fusion, the method also proposes and exploits a simple, but efficient image processing-based algorithm for identifying flat terrain areas (under the assumption that people do not gather on highly curved or inclined terrain), relying on a priori available ground elevation data for the mapped area. Evaluation on both synthetic and real-world multiview video sequences depicting human crowds in outdoor environments verifies the effectiveness of the proposed method.

3D pointing gestures as target selection tools: guiding monocular UAVs during window selection in an outdoor environment

Firefighters need to gain information from both inside and outside of buildings in first response emergency scenarios. For this purpose, drones are beneficial. This paper presents an elicitation study that showed firefighters’ desires to collaborate with autonomous drones. We developed a Human–Drone interaction (HDI) method for indicating a target to a drone using 3D pointing gestures estimated solely from a monocular camera. The participant first points to a window without using any wearable or body-attached device. Through the drone’s front-facing camera, the drone detects the gesture and computes the target window. This work includes a description of the process for choosing the gesture, detecting and localizing objects, and carrying out the transformations between coordinate systems. Our proposed 3D pointing gesture interface improves on 2D interfaces by integrating depth information with SLAM and solving ambiguity with multiple objects aligned on the same plane in a large-scale outdoor environment. Experimental results showed that our 3D pointing gesture interface obtained average F1 scores of 0.85 and 0.73 for precision and recall in simulation and real-world experiments and an F1 score of 0.58 at the maximum distance of 25 m between the drone and building.

An Integrated GNSS/LiDAR-SLAM Pose Estimation Framework for Large-Scale Map Building in Partially GNSS-Denied Environments

This article presents an integrated global navigation satellite system/light detection and ranging (GNSS/LiDAR)-based simultaneous localization and mapping (SLAM) pose estimation framework to perform large-scale 3-D map building in partially GNSS-denied outdoor environments. The framework takes the advantage of the complementarity between GNSS positioning and LiDAR-SLAM to decompose the map building task according to the GNSS real-time kinematic (RTK) status. When mapping in GNSS-denied scenes, a 3-D LiDAR-SLAM algorithm is adopted to estimate poses and a correction algorithm is presented to correct drift errors. On the other hand, when mapping in open scenes, a GNSS-initialized LiDAR mapping algorithm (GL-mapping) is proposed to loosely couple GNSS positioning and LiDAR data registration. It can perform the orientation estimation without the use of either the high-cost inertial sensing device or the GNSS dual-antenna. Experiments are conducted in large-scale outdoor environments to demonstrate that the proposed framework can accomplish simultaneous pose estimation and map building with high precision in both open scenes and GNSS-denied scenes.

VIMO: A Visual-Inertial-Magnetic Navigation System Based on Non-Linear Optimization.

Visual-inertial navigation systems are credited with superiority over both pure visual approaches and filtering ones. In spite of the high precision many state-of-the-art schemes have attained, yaw remains unobservable in those systems all the same. More accurate yaw estimation not only means more accurate attitude calculation but also leads to better position estimation. This paper presents a novel scheme that combines visual and inertial measurements as well as magnetic information for suppressing deviation in yaw. A novel method for initializing visual-inertial-magnetic odometers, which recovers the directions of magnetic north and gravity, the visual scalar factor, inertial measurement unit (IMU) biases etc., has been conceived, implemented, and validated. Based on non-linear optimization, a magnetometer cost function is incorporated into the overall optimization objective function as a yawing constraint among others. We have done extensive research and collected several datasets recorded in large-scale outdoor environments to certify the proposed system’s viability, robustness, and performance. Cogent experiments and quantitative comparisons corroborate the merits of the proposed scheme and the desired effect of the involvement of magnetic information on the overall performance.

Real-Time Robust Stereo Visual SLAM System Based on Bionic Eyes

Visual Simultaneously Localization and Mapping (SLAM) systems obtain accurate ego-motion of cameras and 3D representation of the environment. Although the performance of SLAM systems is very impressive in many common scenes, tracking failure is still a very challenging issue, which always exist in low-textured environment and rapid camera motion situations. In this paper, we propose a real-time stereo visual SLAM system using bionic eyes. The new type sensor named bionic eyes are basically the stereo cameras and perform all the movements of human eyes. Our system makes a good combination of sensors and algorithms. In order to find more feature points, we proposed an autonomous searching strategy of the bionic eyes, which is mainly inspired by the peripheral and central visual of human eyes. After that the camera motion is calculated using a robust motion estimation method. Our system is tested in indoor low-textured and large-scale outdoor environment. Experimental results show that compared to fixed stereo cameras, our bionic eyes SLAM system gains more robustness and avoid tracking failure problem when facing the low-textured environment.

Gender differences in children's wayfinding

ABSTRACT Gender differences in spatial ability are widely acknowledged, but research to date is equivocal on gender differences in wayfinding ability of children. Indeed, the precise nature of the relationship between gender and wayfinding needs further exploration, especially within real-world outdoor environments. There is a need for more research on this topic, not least to better understand the implications of gender differences in wayfinding for facilitation of our educational programs. This article focuses on this issue by seeking to explore children's map understanding and map use in relation to gender in a large-scale outdoor environment. Ninety-seven children, 3–13 years old, were presented with a pictorial map with 7 controls which they had to locate in any order. Measure of accuracy in wayfinding (number of correctly identified controls) indicate small gender differences. However, measure of efficiency in wayfinding (time spent on the task) indicate that boys outperform girls.

3D Registration and Integrated Segmentation Framework for Heterogeneous Unmanned Robotic Systems

The paper proposes a novel framework for registering and segmenting 3D point clouds of large-scale natural terrain and complex environments coming from a multisensor heterogeneous robotics system, consisting of unmanned aerial and ground vehicles. This framework involves data acquisition and pre-processing, 3D heterogeneous registration and integrated multi-sensor based segmentation modules. The first module provides robust and accurate homogeneous registrations of 3D environmental models based on sensors’ measurements acquired from the ground (UGV) and aerial (UAV) robots. For 3D UGV registration, we proposed a novel local minima escape ICP (LME-ICP) method, which is based on the well known iterative closest point (ICP) algorithm extending it by the introduction of our local minima estimation and local minima escape mechanisms. It did not require any prior known pose estimation information acquired from sensing systems like odometry, global positioning system (GPS), or inertial measurement units (IMU). The 3D UAV registration has been performed using the Structure from Motion (SfM) approach. In order to improve and speed up the process of outliers removal for large-scale outdoor environments, we introduced the Fast Cluster Statistical Outlier Removal (FCSOR) method. This method was used to filter out the noise and to downsample the input data, which will spare computational and memory resources for further processing steps. Then, we co-registered a point cloud acquired from a laser ranger (UGV) and a point cloud generated from images (UAV) generated by the SfM method. The 3D heterogeneous module consists of a semi-automated 3D scan registration system, developed with the aim to overcome the shortcomings of the existing fully automated 3D registration approaches. This semi-automated registration system is based on the novel Scale Invariant Registration Method (SIRM). The SIRM provides the initial scaling between two heterogenous point clouds and provides an adaptive mechanism for tuning the mean scale, based on the difference between two consecutive estimated point clouds’ alignment error values. Once aligned, the resulting homogeneous ground-aerial point cloud is further processed by a segmentation module. For this purpose, we have proposed a system for integrated multi-sensor based segmentation of 3D point clouds. This system followed a two steps sequence: ground-object segmentation and color-based region-growing segmentation. The experimental validation of the proposed 3D heterogeneous registration and integrated segmentation framework was performed on large-scale datasets representing unstructured outdoor environments, demonstrating the potential and benefits of the proposed semi-automated 3D registration system in real-world environments.

Sparse semantic map building and relocalization for UGV using 3D point clouds in outdoor environments

In this paper, we proposed a sparse semantic map building method and an outdoor relocalization strategy based on this map. Most existing semantic mapping approaches focus on improving semantic understanding of single frames and retain a large amount of environmental data. Instead, we don't want to locate the UGV precisely, but use the imprecise environmental information to determine the general position of UGV in a large-scale environment like human beings. For this purpose, we divide the environment into environment nodes according to the result of scene understanding. The semantic map of the outdoor environment is obtained by generating topological relations between the environment nodes. In the semantic map, only the information of the nodes is saved, so that the storage space can be kept at a very small level with the increasing size of environment. When the UGV receives a new local semantic map, we evaluate the similarity between local map and global map to determine the possible position of the UGV according to the categories of the left and right nodes and the distance between the current position and the nodes. In order to validate the proposed approach, experiments have been conducted in a large-scale outdoor environment with a real UGV. Depending on the semantic map, the UGV can redefine its position from different starting points.

Indirect Visual Simultaneous Localization and Mapping Based on Linear Models

This paper proposes an indirect visual simultaneous localization and mapping (V-SLAM) for large-scale outdoor environments based on linear models. With the scale-invariant feature transform (SIFT) algorithm, a design combining key-frame insertion and map management is proposed to avoid redundant computation and unreliable landmarks. Then, an iterative linear equation is employed to update landmarks. With the supplement of reliable map data, absolute camera poses can be obtained using a simple linear equation based on a linearity index. In addition, a linear model helps the system to detect candidates of looped frames and discard invalid loops by means of an outlier weight function. If a loop is detected, an improved trajectory bending algorithm is introduced to optimize the estimations of poses and landmarks. To evaluate the effectiveness of the proposed approach, extensive experiments are conducted using various sequences from a well-known public dataset. Experimental results show that the proposed V-SLAM outperforms other existing methods in terms of various metrics.

Deep-Learning-based Relocalization in Large-Scale outdoor Environment

Abstract For the issue of relocalization, this paper proposed a deep-learning-based method for outdoor large-scale environment. In the first step, we projected a 3D Light Detection and Ranging(LiDAR) scan onto three 2D images from top to bottom. Then a densenet-based neural network structure was designed to regress a 4-DOF robot pose. These images are then stacked together, fed into the proposed DCNN architecture, and the output is the predicted robot pose. Extensive experiments have been conducted in practice with a real mobile robot, verifying the effectiveness of the proposed strategy. Our network can obtain approximately 3.5m and 4◦ accuracy outdoors.

Panoramic SLAM from a multiple fisheye camera rig

This paper presents a feature-based simultaneous localization and mapping (SLAM) system for panoramic image sequences obtained from a multiple fisheye camera rig in a wide baseline mobile mapping system (MMS). First, the developed fisheye camera calibration method combines an equidistance projection model and trigonometric polynomial to achieve high-accuracy calibration from the fisheye camera to an equivalent ideal frame camera, which warrants an accurate transform from the fisheye images to a corresponding panoramic image. Second, we developed a panoramic camera model, corresponding bundle adjustment with a specific back propagation error function, and linear pose initialization algorithm. Third, the implemented feature-based SLAM pipeline consists of several specific strategies and algorithms in initialization, feature matching, frame tracking, and loop closing to overcome the difficulties in tracking wide baseline panoramic image sequences. We conducted experiments on large-scale MMS datasets of more than 15 km trajectories and 14,000 panoramic images as well as small-scale public video datasets. Our results show that the developed panoramic SLAM system PAN-SLAM can achieve fully-automatic camera localization and sparse map reconstruction in both small-scale indoor and large-scale outdoor environments including challenging scenes (e.g., dark tunnel) without the aid of any other sensors. The localization accuracy, which was measured by the absolute trajectory error (ATE), resembled the high-accuracy GNSS/INS reference of 0.1 m. PAN-SLAM also outperformed several feature-based fisheye and monocular SLAM systems with incomparable robustness in various environments. The system could be considered as a robust complementary solution and an alternative to expensive commercial navigation systems, especially in urban environments where signal obstruction and multipath interference are common. Source code and demo are available at http://study.rsgis.whu.edu.cn/pages/download/.