Building Visual Maps with a Team of Mobile Robots

Building maps is one of the most fundamental tasks for an autonomous robot. This robot should be able to construct a map of the environment and, at the same time, localize itself in it. This problem, known as Simultaneous Localization and Mapping (SLAM), has received great interest in the last years (Leonard & Durrant-Whyte, 1991). In our particular case, the robots build their maps using the FastSLAM algorithm (Montemerlo et al., 2002). The main idea of the FastSLAM algorithm is to separate the two fundamental aspects of the SLAM problem: the estimate of the robot’s pose and the estimate of the map. This algorithm uses a particle set that represents the uncertainty of the robot’s pose (localization problem) meanwhile each particle has its own associated map (several individual estimates of the landmarks conditioned to the robot’s path). The solution to the SLAM problem is performed by means of a sampling and particle generation process, in which the particles whose current observations do not fit with their associated map are eliminated. The FastSLAM algorithm has proved to be robust to false data association and it is able to represent models of non-linear movements in a reliable way (Howard, 2006). In relation to the sensors used to build the maps, many authors use range sensors such as sonar (Kwak et al., 2008; Wijk & Christensen, 2000) or laser (Thrun, 2004; Triebel & Burgard, 2005). Nevertheless, there is an increasing interest on using cameras as sensors. This approach is denoted as visual SLAM (Valls Miro et al., 2006). These devices obtain a higher amount of information from the environment and they are less expensive than laser as well. Furthermore, 3D information can be obtained when using stereo cameras. These are the sensors used in this work. Most visual SLAM approaches are landmark-based. These landmarks consist of a set of distinctive points which are referred to a global reference system. The main advantage of landmark-based maps is the compactness of their representation. By contrast, this kind of maps requires the existence of structures or objects that are distinguishable enough. The map building problem can be solved by a single robot (Moutalier & Chatila, 1989), but it will be more efficiently managed if there is a team of robots, which collaborates in the construction of the map (Howard, 2006). In this case, the space can be divided so that the distances traversed are shorter and thus the odometry errors will be smaller. In this work, 7

Teams of mobile robots can be deployed in search and rescue missions to explore previously unknown environments. Methods for joint localization and mapping constitute the basis for (semi-)autonomous cooperative action, in particular when navigating in GPS-denied areas. As communication losses may occur, a decentralized solution is required. With these challenges in mind, we designed a submap-based SLAM system that relies on inertial measurements and stereo-vision to create multi-robot dense 3D maps. For online pose and map estimation, we integrate the results of keyframe-based local reference filters through incremental graph SLAM. To the best of our knowledge, we are the first to combine these two methods to benefit from their particular advantages for 6D multi-robot localization and mapping: Local reference filters on each robot provide real-time, long-term stable state estimates that are required for stabilization, control and fast obstacle avoidance, whereas online graph optimization provides global multi-robot pose and map estimates needed for cooperative planning. We propose a novel graph topology for a decoupled integration of local filter estimates from multiple robots into a SLAM graph according to the filters' uncertainty estimates and independence assumptions and evaluated its benefits on two different robots in indoor, outdoor and mixed scenarios. Further, we performed two extended experiments in a multi-robot setup to evaluate the full SLAM system, including visual robot detections and submap matches as inter-robot loop closure constraints.

Multi-camera visual SLAM for autonomous navigation of micro aerial vehicles

A comprehensive overview of core modules in visual SLAM framework

Visual simultaneous localization and mapping (VSLAM) is an essential technique used in areas such as robotics and augmented reality for pose estimation and 3D mapping. Research on VSLAM using both monocular and stereo cameras has grown significantly over the last two decades. There is, therefore, a need for emphasis on a comprehensive review of the evolving architecture of such algorithms in the literature. Although VSLAM algorithm pipelines share similar mathematical backbones, their implementations are individualized and the ad hoc nature of the interfacing between different modules of VSLAM pipelines complicates code reuseability and maintenance. This paper presents a software model for core components of VSLAM implementations and interfaces that govern data flow between them while also attempting to preserve the elements that offer performance improvements over the evolution of VSLAM architectures. The framework presented in this paper employs principles from model-driven engineering (MDE), which are used extensively in the development of large and complicated software systems. The presented VSLAM framework will assist researchers in improving the performance of individual modules of VSLAM while not having to spend time on system integration of those modules into VSLAM pipelines.

Current visual simultaneous localization and mapping (SLAM) systems have demonstrated commendable efficacy in static environments. However, the presence of dynamic objects in real-world settings frequently leads to system discrepancies, significantly impairing the accuracy and robustness of SLAM systems. Conventional visual SLAM approaches typically utilize epipolar constraints to mitigate the impact of outliers; nevertheless, they encounter limitations when confronted with a substantial number of dynamic or planar moving objects. To tackle these issues, this paper introduces a novel dynamic visual SLAM system, termed DEG-SLAM. Initially, the system employs the YOLOv5 object detection network to identify dynamic objects, subsequently relaying the semantic information to the tracking module. During the tracking phase, both semantic information and epipolar constraints are leveraged to filter out dynamic feature points. To address the challenges posed by the malfunctioning of epipolar constraints in degenerate scenes, DEG-SLAM incorporates a degenerate constraint mechanism aimed at further eliminating dynamic feature points. Furthermore, a reprojection constraint has been introduced to enhance the filtering of absent dynamic feature points that lie outside the detection boxes. Experimental findings reveal that DEG-SLAM significantly improves accuracy and robustness when compared to traditional ORB-SLAM3 in dynamic environments. The performance benefits of DEG-SLAM are particularly evident in degenerate scenarios, thereby affirming its practicality and reliability in complex settings.

Whenever multiple robots have to solve a common task, they need to coordinate their actions to carry out the task efficiently and to avoid interferences between individual robots. This is especially the case when considering the problem of exploring an unknown environment with a team of mobile robots. To achieve efficient terrain coverage with the sensors of the robots, one first needs to identify unknown areas in the environment. Second, one has to assign target locations to the individual robots so that they gather new and relevant information about the environment with their sensors. This assignment should lead to a distribution of the robots over the environment in a way that they avoid redundant work and do not interfere with each other by, for example, blocking their paths. In this paper, we address the problem of efficiently coordinating a large team of mobile robots. To better distribute the robots over the environment and to avoid redundant work, we take into account the type of place a potential target is located in (e.g., a corridor or a room). This knowledge allows us to improve the distribution of robots over the environment compared to approaches lacking this capability. To autonomously determine the type of a place, we apply a classifier learned using the AdaBoost algorithm. The resulting classifier takes laser range data as input and is able to classify the current location with high accuracy. We additionally use a hidden Markov model to consider the spatial dependencies between nearby locations. Our approach to incorporate the information about the type of places in the assignment process has been implemented and tested in different environments. The experiments illustrate that our system effectively distributes the robots over the environment and allows them to accomplish their mission faster compared to approaches that ignore the place labels.

Recently, the problem of map building and simultaneous localization is a hot topic in mobile robotics field In this paper, a feature map building approach based on shape similarity for the development of home robot is presented In the map building process, the robot is continuously localized as the map is being built. The odometry error between each constitutive step is eliminated by comparing the shape similarity of the corresponding objects in scan obtained from the laser ranger finder against the global map, which was built in the previous step. This developed algorithm is capable of producing faithful global map of indoor environment, as well as handling uncertain information, such as moving objects and fuzzy-shaped objects. The experiment is carried out using home robot ApriAttendatrade, developed by Toshiba Corporation.

We consider the task of controlling a large team of nonholonomic ground robots with an unmanned aerial vehicle in a decentralized manner that is invariant to the number of ground robots. The central idea is the development of an abstraction for the team of ground robots that allows the aerial platform to control the team without any knowledge of the specificity of individual vehicles. This happens in much the same way as a human operator can control a single robot vehicle by simply commanding the forward and turning velocities without a detailed knowledge of the specifics of the robot. The abstraction includes a gross model of the shape of the formation of the team and information about the position and orientation of the team in the plane. We derive controllers that allow the team of robots to move in formation while avoiding collisions and respecting the abstraction commanded by the aerial platform. We provide simulation and experimental results using a team of indoor mobile robots and a three-dimensional, cable-controlled, parallel robot which serves as our indoor unmanned aerial platform.

With the development of novel RGB-D visual sensors, data association has been a basic problem in 3D Visual Simultaneous Localization and Mapping (VSLAM). To solve the problem, a VSLAM algorithm based on Multiple Iterative Closest Point (MICP) is presented. By using both RGB and depth information obtained from RGB-D camera, 3D models of indoor environment can be reconstructed, which provide extensive knowledge for mobile robots to accomplish tasks such as VSLAM and Human-Robot Interaction. Due to the limited views of RGB-D camera, additional information about the camera pose is needed. In this paper, the motion of the RGB-D camera is estimated by a motion capture system after a calibration process. Based on the estimated pose, the MICP algorithm is used to improve the alignment. A Kinect mobile robot which is running Robot Operating System and the motion capture system has been used for experiments. Experiment results show that not only the proposed VSLAM algorithm achieved good accuracy and reliability, but also the 3D map can be generated in real time.

Interest point extraction is a crucial technique for visual simultaneous localization and mapping, which can affect the accuracy of the whole system's localization. Feature point extraction is closely related to image quality. In real life, the motion blur phenomenon is always unavoidable in the process of camera shooting. For example, the fast motion of the camera and the rapid movement of the object will cause the motion blur of the image, which will lead to the inaccurate extraction location of interest points. In this paper we propose a DeblurSLAM framework, which is a Visual SLAM system building on ORB-SLAM2 [1]. DeblurSLAM is robust in blurring scenarios for RGB-D and stereo configurations. We are capable of detecting the blur and removing blur interference. To our knowledge, it is the first work combining the deblurring network into a Visual SLAM system. We conduct experiments both on TUM RGB-D and KITTI stereo datasets. The results demonstrate that DeblurSLAM outperforms the accuracy of standard Visual SLAM baselines.

This paper presents a new method to perform collaborative real-time dense 3D mapping in a distributed way for a multi-robot system. This method associates a Truncated Signed Distance Function (TSDF) representation with a manifold structure. Each robot owns a private map which is composed of a collection of local TSDF sub-maps called patches that are locally consistent. This private map can be shared to build a public map collecting all the patches created by the robots of the fleet. In order to maintain consistency in the global map, a mechanism of patch alignment and fusion has been added. This work has been integrated in real-time into a mapping stack, which can be used for autonomous navigation in unknown and cluttered environment. Experimental results on a team of wheeled mobile robots are reported to demonstrate the practical interest of the proposed system, in particular for the exploration of unknown areas.

Building Information Modeling (BIM) has been widely adopted for construction inspections due to its ability to integrate multiple data sources. Engineers use BIM to identify and review site issues, yet inspection systems face several challenges. Firstly, positioning inspection areas on a construction site using BIM with Augmented Reality (AR) requires complex model manipulation. Additionally, signal or Internet connectivity issues may limit positioning technologies. Secondly, human error or interference is common in traditional inspection processes due to their complexity. To overcome these barriers, this research applied BIM and AR with Visual Simultaneous Localization and Mapping (VSLAM) to help inspectors quickly and effectively record construction defects as photographs with notes and their locations. An efficient approach is proposed to integrate BIM and AR with VSLAM, and a prototype is developed to validate and demonstrate how the proposed system can assist a site inspector in performing quality management, even offline. The system uses a two-phase indoor positioning method: initial localization via visual markers and real-time tracking with VSLAM, enabling precise defect tracking and efficient model adjustments. While significantly improving inspection accuracy and efficiency, its performance is affected by environmental factors like lighting and marker placement, providing insights for future refinement.

Visual simultaneous localization and mapping (VSLAM) is extensively utilized in the field of unmanned driving. Classical VSLAM typically postulates that the surroundings are stationary, and dynamic objects can seriously affect the accuracy of the system. However, detecting and tracking dynamic objects is beneficial for improving the autonomy and intelligence of robotics. In order to solve the problem of VSLAM distinguishing dynamic objects in dynamic scenes, a semantic VSLAM system based on the VDO-SLAM framework is proposed. The approach integrates semantic segmentation into the visual odometry and combines scene flow and geometric constraints to effectively reduce the negative impact of dynamic objects on system performance. Specifically, a transformer-based neural network architecture for optical flow estimation was employed, which effectively enhances the segmentation precision of foreground and background regions. And a geometric constraint was introduced to classify dynamic objects, which can reduction of erroneous segmentation of dynamic objects with effect. The data analysis results on the KITTI dataset demonstrate that our proposed algorithms can effectively enhance precision and stability of VSLAM system, reduce the tracking error of VSLAM in dynamic environments, and demonstrate good potential and application value.

Visual Simultaneous Localization and Mapping (VSLAM) is a critical technology for intelligent mobile robots, enabling simultaneous environmental mapping and self-localization. However, reflective ground in indoor environments presents significant challenges to VSLAM systems, as it introduces noisy virtual landmarks that degrade both map accuracy and localization precision. To address these challenges, we propose a novel RGB-D VSLAM system specifically designed for these environments. Our approach begins with a carefully designed point-line-plane-based VSLAM framework, which mitigates the limitations caused by sparse point features in indoor settings. Then we analyze how reflective ground degrades VSLAM performance and propose a reflective ground feature removal strategy that integrates semantic and geometric information. To enhance robustness in highly reflective environments, we incorporate a monocular depth estimation network as a complementary module. Extensive experiments on public datasets demonstrate that our framework achieves state-of-the-art performance. And evaluations on an author-collected dataset highlight the system’s superior mapping and localization capabilities in reflective indoor environments. Furthermore, ablation studies validate the effectiveness of each proposed component, and latency tests confirm the system’s suitability for real-time applications.