SLAM Method Research Articles

Tracking fiducial markers with discriminative correlation filters

In the last few years, squared fiducial markers have become a popular and efficient tool to solve monocular localization and tracking problems at a very low cost. Nevertheless, marker detection is affected by noise and blur: small camera movements may cause image blurriness that prevents marker detection.The contribution of this paper is two-fold. First, it proposes a novel approach for estimating the location of markers in images using a set of Discriminative Correlation Filters (DCF). The proposed method outperforms state-of-the-art methods for marker detection and standard DCFs in terms of speed, precision, and sensitivity. Our method is robust to blur and scales very well with image resolution, obtaining more than 200fps in HD images using a single CPU thread.As a second contribution, this paper proposes a method for camera localization with marker maps employing a predictive approach to detect visible markers with high precision, speed, and robustness to blurriness. The method has been compared to the state-of-the-art SLAM methods obtaining, better accuracy, sensitivity, and speed. The proposed approach is publicly available as part of the ArUco library.

A Novel 3D LiDAR SLAM Based on Directed Geometry Point and Sparse Frame

Simultaneous localization and mapping is an indispensable yet challenging direction for mobile robots. Attracted by 3D LiDAR with accurate depth information and robustness to illumination variations, many 3D LiDAR SLAM methods based on scan-to-map matching have been developed. However, there is a critical issue of existing approaches, where a large and dense map is generally required to achieve satisfactory localization accuracy, leading to low efficiency of scan-to-map matching. To solve this problem, in this letter, we propose a novel 3D LiDAR SLAM based on directed geometry point (DGP) and sparse frame. The former is used to provide a sparse distribution of points in the spatial dimension and the latter gives rise to a sparse distribution of frames in the temporal sequence. The sparsity of points and frames impove the efficiency of 3D LiDAR SLAM, and the strict data association based on directed geometric points also brings in good accuracy of pose estimation. To compensate the accuracy loss of the localization and mapping caused by frame sparsity, point propagation is proposed to improve the quality of directed geometric points in the map and the accuracy of scan-to-map matching. Also, loop detection and pose graph optimization are conducted for global consistency. The experimental results demonstrate the effectiveness of the proposed method in terms of accuracy and efficiency.

An improved ORB-SLAM2 algorithm based on image information entropy

Aiming at the feature point extraction process in visual ORB-SLAM2, a visual SLAM method based on information entropy is proposed. In the process of ORB-SLAM2, due to the uncertainty of illumination and surrounding environment information, when the environment information is reduced, the number of feature points is likely to decrease dramatically, which results in the failure of SLAM process. In order to solve the above problems, a visual SLAM method based on image information entropy is proposed. According to the information entropy of the image, the amount of information is determined, and the image with low contrast and small gradient change is screened. The image is enhanced, and the feature points that can represent the image information are extracted as much as possible as the correlation basis of adjacent frame matching and key frame matching. In order to improve the performance of the mobile robot system, the local and global optimization of cumulative error of position and attitude is carried out by combining the optimization method of attitude graph.

Structure-SLAM: Low-Drift Monocular SLAM in Indoor Environments

In this letter a low-drift monocular SLAM method is proposed targeting indoor scenarios, where monocular SLAM often fails due to the lack of textured surfaces. Our approach decouples rotation and translation estimation of the tracking process to reduce the long-term drift in indoor environments. In order to take full advantage of the available geometric information in the scene, surface normals are predicted by a convolutional neural network from each input RGB image in real-time. First, a drift-free rotation is estimated based on lines and surface normals using spherical mean-shift clustering, leveraging the weak Manhattan World assumption. Then translation is computed from point and line features. Finally, the estimated poses are refined with a map-to-frame optimization strategy. The proposed method outperforms the state of the art on common SLAM benchmarks such as ICL-NUIM and TUM RGB-D.

Monocular Visual SLAM for Underwater Navigation in Turbid and Dynamic Environments

Localization, navigation, and mapping using vision-based algorithms are an active topic in underwater robotic applications. Although many algorithms developed in recent years, especially in the ground and areal robotic communities, directly applying those methods in underwater navigation remain challenging due to the visual degradation induced by the medium. In this paper, we proposed UW-SLAM (Underwater SLAM), a new monocular visual SLAM algorithm focused on the underwater environment which addresses the turbidity and dynamism. The proposed method was evaluated with several underwater datasets with comparison to the state of the art monocular SLAM methods.

From Simultaneous Localization and Mapping to Collaborative Localization for Intelligent Vehicles

This paper describes a new method for real-time collaborative localization for intelligent vehicles based on a combination of an extension of simultaneous localization and mapping and track-to-track fusion methods. This strategy aims to overcome the shortcomings of the embedded proprioceptive sensors and solve the localization issues in a cluster of intelligent vehicles, which are equipped with low-cost navigation systems (global position systems, rangefinders, and vehicle to vehicle communication means) in an unknown environment. The strategy improves the vehicles’ localization accuracy and tracks a cluster of vehicles for advanced driver assistance systems and vehicles self-driving. For real-time purposes, the computational complexity of standard solutions, which are based on Kalman filter derivatives, is transformed, herein, from cubic to quadratic with respect to the state vector of the cluster. Furthermore, the correlated data is another challenge that needs to be solved in cooperative localization scenarios. To this end, two independent, local and global, fusion nodes are associated with each vehicle. The local node exploits multiple forms of the extended Kalman filter and its dual extended information filter to estimate the vehicles’ state, information vectors, covariance and information matrices in the moments and information forms, respectively. Then, the vehicles broadcast the result of their respective local fusion node to cluster. The information is used to solve the second problem, which is a track-to-track fusion on distributed architectures, with two methods: covariance intersection and information matrix fusion in the global fusion node. The simulation and the experimental results of the proposed methods, on distributed architectures, are compared to the optimal solution of centralized data fusion, and to the traditional SLAM method in terms of root mean square errors and consistency.

Multi-Feature Nonlinear Optimization Motion Estimation Based on RGB-D and Inertial Fusion

To achieve a high precision estimation of indoor robot motion, a tightly coupled RGB-D visual-inertial SLAM system is proposed herein based on multiple features. Most of the traditional visual SLAM methods only rely on points for feature matching and they often underperform in low textured scenes. Besides point features, line segments can also provide geometrical structure information of the environment. This paper utilized both points and lines in low-textured scenes to increase the robustness of RGB-D SLAM system. In addition, we implemented a fast initialization process based on the RGB-D camera to improve the real-time performance of the proposed system and designed a new backend nonlinear optimization framework. By minimizing the cost function formed by the pre-integrated IMU residuals and re-projection errors of points and lines in sliding windows, the state vector is optimized. The experiments evaluated on public datasets show that our system achieves higher accuracy and robustness on trajectories and in pose estimation compared with several state-of-the-art visual SLAM systems.

A Distributed Approach for the Implementation of Geometric Reconstruction-Based Visual SLAM Systems

SUMMARYVisual simultaneous localization and mapping (VSLAM) is a relevant solution for vehicle localization and mapping environments. However, it is computationally expensive because it demands large computational effort, making it a non-real-time solution. The VSLAM systems that employ geometric reconstructions are based on the parallel processing paradigm developed in the Parallel Tracking and Mapping (PTAM) algorithm. This type of system was created for processors that have exactly two cores. The various SLAM methods based on the PTAM were also not designed to scale to all the cores of modern processors nor to function as a distributed system. Therefore, we propose a modification to the pipeline for the execution of well-known VSLAM systems so that they can be scaled to all available processors during execution, thereby increasing their performance in terms of processing time. We explain the principles behind this modification via a study of the threads in the SLAM systems based on PTAM. We validate our results with experiments describing the behavior of the original ORB-SLAM system and the modified version.

2D Laser SLAM With General Features Represented by Implicit Functions

The main contribution of this letter is the problem formulation and algorithm framework for 2D laser SLAM with general features represented by implicit functions. Since 2D laser data reflect the distances from the robot to the boundary of objects in the environment, it is natural to use the boundary of the general objects/features within the 2D environment to describe the features. Implicit functions can be used to represent almost arbitrary shapes from simple (e.g. circle, ellipse, line) to complex (e.g. a cross-section of a bunny model), thus it is worth studying implicit-expressed feature in 2D laser SLAM. In this letter, we clearly formulate the SLAM problem with implicit functions as features, with rigorously computed observation covariance matrix to be used in the SLAM objective function and propose a solution framework. Furthermore, we use ellipses and lines as examples to compare the proposed SLAM method with the traditional pre-fit method (represent the feature using its parameters and pre-fit the laser scan to get the fitted parameter as virtual observations). Simulation and experimental results show that our proposed method has a better performance compared with the pre-fit method and other methods, demonstrating the potential of this new SLAM formulation and method.

Optimizing Strap-down Inertial Navigation Systems Characteristics and External Correction Sensors for Autonomous Navigation of Different UAV Classes

Currently, unmanned aerial vehicles (UAVs) can be used in topographic works, condition monitoring and diagnostics of extended engineering structures, delivering goods to hard-to-reach places, etc. To provide the widespread UAVs applications and raise the number of tasks to be solved through their using, it is necessary to increase their autonomy degree in terms of navigation support, in particular. Unmanned aerial vehicles (UAV) control systems for autonomous navigation use the strap-down inertial navigation systems (SINS) based on various types of gyroscopes. SINS based on the laser gyroscopes, which have a large mass, have the best accuracy. UAVs with a payload mass that is commensurable with the mass of navigation equipment require optimization of SINS characteristics. An optimization method has been developed to enable obtaining a Pareto set for the mass and accuracy of SINS based on laser gyroscopes. A comprehensive assessment of the characteristics of SINS and UAV carrier with different payload mass has been performed. Various SINS correction methods are considered when satellite navigation is unavailable.For overland flights, the correlation-extreme navigation systems (CENS) and SLAM methods (for simultaneous localisation and mapping) can be used. CENS require a reference lay-of-the-land description and a sufficient density of landmarks. In navigation based on SLAM algorithms, there is no need in the reference lay-of-the-land description, and the initial data can be obtained through the optical sensors under appropriate condition of the atmospheric path.Regardless of the condition of the atmospheric path, type of the underlying surface and its information available in detail, the UAV coordinates can be determined by Doppler dead reckoning using a Doppler system (DISS). At low and medium altitudes SINS correction is possible, only heading sensor data are needed to calculate the path angle.In combining with DISS and 3D Flash Ladar sensors (for implementing SLAM algorithms), it is more optimal to use low-accuracy SINS based on fibre-optic gyroscopes rather than laser gyro-based systems.The results obtained can be used when developing navigation systems for medium, light and heavy-medium UAVs.

PanoraMIS: An ultra-wide field of view image dataset for vision-based robot-motion estimation

This article presents a new dataset of ultra-wide field of view images with accurate ground truth, called PanoraMIS. The dataset covers a large spectrum of panoramic cameras (catadioptric, twin-fisheye), robotic platforms (wheeled, aerial, and industrial robots), and testing environments (indoors and outdoors), and it is well suited to rigorously validate novel image-based robot-motion estimation algorithms, including visual odometry, visual SLAM, and deep learning-based methods. PanoraMIS and the accompanying documentation is publicly available on the Internet for the entire research community.

Research on Target-Driven Navigation of Mobile Robot Based on Deep Reinforcement Learning and Preprocessing Layer

Recently, with the rise of deep reinforcement learning model, robot navigation based on this method has a huge advantage compared with traditional slam method, which has attracted extensive attention. However, when the navigation algorithm trained in the virtual environment is transferred to the real environment, the navigation performance of the robot will decline sharply because of the great difference between the virtual environment and the real environment. In order to improve the navigation ability of mobile robot, this paper implements a mobile robot navigation system based on deep reinforcement learning without environment map and only visual input. At the same time, in order to solve the problem of poor generalization ability of deep reinforcement learning from virtual environment to real environment, this paper proposes a preprocessing layer with knowledge and combines it with deep reinforcement learning module. The combined algorithm model alleviates the performance fault problem caused by the migration algorithm and the performance difference between virtual sensor and real sensor. At the end of this paper, a navigation experiment based on the turtlebot is designed, which proves that the deep reinforcement learning algorithm with the preprocessing layer can alleviate the performance fault problem caused by the migration algorithm, and have a certain ability of obstacle avoidance and avoidance without the environment map.

GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization

Direct SLAM methods have shown exceptional performance on odometry tasks. However, they are susceptible to dynamic lighting and weather changes while also suffering from a bad initialization on large baselines. To overcome this, we propose GN-Net: a network optimized with the novel Gauss-Newton loss for training weather invariant deep features, tailored for direct image alignment. Our network can be trained with pixel correspondences between images taken from different sequences. Experiments on both simulated and real-world datasets demonstrate that our approach is more robust against bad initialization, variations in daytime, and weather changes thereby outperforming state-of-the-art direct and indirect methods. Furthermore, we release an evaluation benchmark for relocalization tracking against different types of weather. Our benchmark is available at https://vision.in.tum.de/ gn-net.

Exploring the spectral information content in the LAMOST medium-resolution survey (MRS)

Although high-resolution stellar spectra allow us to derive precise stellar labels (effective temperature, metallicity, surface gravity, elemental abundances, etc.) based on resolved atomic lines and molecular bands, low-resolution spectra have been proved to be competitive in determiningmany stellar labels at comparable precision. It is useful to consider the spectral information content when assessing the capability of a stellar spectrum in deriving precise stellar labels. In this work, we quantify the information content brought by the LAMOST-II medium-resolution spectroscopic survey (MRS) using the gradient spectra as well as the coefficients-of-dependence (CODs). In general, the wavelength coverage of the MRS well constrains the stellar labels but the sensitivities of different stellar labels vary with spectral types and metallicity of the stars of interest. Consequently, this affects the performance of the stellar label determination from the MRS spectra. By applying the SLAM method to the synthetic spectra which mimic the MRS data, we find that the precision of the fundamental stellar parameters Teff, log g and [M/H] are better when combining both the blue and red bands of the MRS. This is especially important for warm stars because the Hα line located in the red part plays a more important role in determining the effective temperature for warm stars. With blue and red parts together, we are able to reach similar performance to the low-resolution spectra except for warm stars. However, at [M/H] ∼ −2.0 dex, the uncertainties of fundamental stellar labels estimated from MRS are substantially larger than that from low-resolution spectra. We also tested the uncertainties of Teff, log g and [M/H] From MRS data induced from the radial velocity mismatch and find that a mismatch of about 1 km s−1, which is typical for LAMOST MRS data, would not significantly affect the stellar label estimates. Finally, reference precision limits are calculated using synthetic gradient spectra, according to which we expect abundances of at least 17 elements to be measured precisely from MRS spectra.

Multi-camera visual SLAM for off-road navigation

With the rapid development of computer vision, vision-based simultaneous localization and mapping (vSLAM) plays an increasingly important role in the field of unmanned driving. However, traditional SLAM methods based on a monocular camera only perform well in simple indoor environments or urban environments with obvious structural features. In off-road environments, the situation that SLAM encounters could be complicated by problems such as direct sunlight, leaf occlusion, rough roads, sensor failure, sparsity of stably trackable texture. Traditional methods are highly susceptible to these factors, which lead to compromised stability and reliability. To counter such problems, we propose a panoramic vision SLAM method based on multi-camera collaboration, aiming at utilizing the characters of panoramic vision and stereo perception to improve the localization precision in off-road environments. At the same time, the independence and information sharing of each camera in multi-camera system can improve its ability to resist bumps, illumination, occlusion and sparse texture in an off-road environment, and enable our method to recover the metric scale. These characters ensure unmanned ground vehicles (UGVs) to locate and navigate safely and reliably in complex off-road environments.

DRM-SLAM: Towards dense reconstruction of monocular SLAM with scene depth fusion

Monocular visual SLAM methods can accurately track the camera pose and infer the scene structure by building sparse correspondence between two/multiple views of the scene. However, the reconstructed 3D maps of these methods are extremely sparse. On the other hand, deep learning is widely used to predict dense depth maps from single-view color images, but the results are subject to blurry depth boundaries, which severely deform the structure of 3D scene. Therefore, this paper proposes a dense reconstruction method under the monocular SLAM framework (DRM-SLAM), in which a novel scene depth fusion scheme is designed to fully utilize both the sparse depth samples from monocular SLAM and predicted dense depth maps from convolutional neural network (CNN). In the scheme, a CNN architecture is carefully designed for robust depth estimation. Besides, our approach also accounts for the problem of scale ambiguity existing in the monocular SLAM. Extensive experiments on benchmark datasets and our captured dataset demonstrate the accuracy and robustness of the proposed DRM-SLAM. The evaluation of runtime and adaptability under challenging environments also verify the practicability of our method.

An hybrid registration method for SLAM with Quanergy M8 LiDAR sensor

An hybrid registration method for SLAM with Quanergy M8 LiDAR sensor

Localization of RFID Tags by a Moving Robot, via Phase Unwrapping and Non-Linear Optimization

In this paper, we propose a new method for the localization of RFID tags, by deploying off-the-shelf RFID equipment on a robotic platform. The constructed robot is capable to perform Simultaneous Localization (of its own position) and Mapping (SLAM) of the environment and then locate the RFID tags around its path. The proposed method is based on properly treating the measured phase of the backscattered signal by each tag at the reader’s antenna, located on top of the robot. More specifically, the measured phase samples are reconstructed, such that the $2\pi $ discontinuities are eliminated (phase-unwrapping). This allows for the formation of an optimization problem, which can be solved rapidly by standard methods. The proposed method is experimentally compared against the SAR/imaging methods, which represent the accuracy benchmark in prior-art, deploying off-the-shelf equipment. It is shown that the proposed method solves exactly the same problem as holographic-imaging methods, overcoming the grid-density constraints of the latter. Furthermore, the problem, being calculations-grid-independent, is solved orders of magnitude faster, allowing for the applicability of the method in real-time inventorying and localization. It is also shown that the state-of-the-art SLAM method, which is used for the estimation of the trace of the robot, also suffers from errors, which directly affect the accuracy of the RFID localization method. Deployment of reference RFID tags at known positions, seems to significantly reduce such errors.

A Novel Texture-Less Object Oriented Visual SLAM System

Traditional mapping modules in Visual Simultaneous Localization and Mapping (i.e. Visual SLAM) systems can only estimate 3D information of isolated sparse or semi-dense feature points. But there are lots of object instances in the environments which geometric information can be utilized to enhance the quality of mapping and localization. Hence, it is required for the Visual SLAM system to utilize high-dimensional features like object instances or structural lines in mapping and localization. To meet the gap between the above requirements and the traditional implementation of Visual SLAM systems, we present in this paper a novel Visual SLAM method that can effectively utilize texture-less object instances for mapping and localization. The proposed Visual SLAM method includes newly designed feature extraction, matching, localization and mapping modules, which jointly use object features and point features to estimate camera 6-DOF poses and do richer map construction. A group of organized raster points is used to represent objects during feature matching and pose estimation process in the proposed Visual SLAM pipeline. Owing to the object feature fusion in the co-visibility graph it could conduct scale aware bundle adjustments to reduce accumulated error. The advantages of proposed Visual SLAM method are demonstrated through experiments conducted both on synthetic datasets and real-world datasets.

A synthetic dataset for Visual SLAM evaluation

Vision-based self-localization methods are key functionalities for various research topics. Recent research results on related fields have catalyzed several accurate, versatile and reliable real-time Visual SLAM systems suitable for self-localization under a wide variety of environmental preconditions. These methods extend their functionalities from being only a good camera tracker to being able to recursively build up camera’s surroundings. The fast development of Visual SLAM research has proposed demands on innovating evaluation methods for Visual SLAM systems. However, retrieving images and ground truth from various kinds of environments, estimating calibration parameters between several sensors and annotating useful labels all require cumbersome human labor and will introduce inevitable errors. In this paper, we propose a method that uses virtually established models to automatically generate photorealistic images with accurate ground truth and several kinds of pixel-level annotations useful for Visual SLAM development and evaluation. We build and render a challenging dataset in low-texture environments with large scale camera movement, multiple moving objects and varying luminance status. We also propose several new evaluation criteria that can fully take advantage of ground truth and annotations from synthetic datasets. Experiments are conducted using the proposed datasets and criteria with several state-of-the-art Visual SLAM methods to demonstrate the functionality of our datasets.