Simultaneous localization and mapping (SLAM) has been widely used in augmented reality (AR), virtual reality (VR), robotics, and autonomous vehicles as the theoretical basis for robots to perceive their environment. Most popular SLAM algorithms assume that objects in the scene are static. Solving dynamic problems in SLAM is now attracting increasing attention. In this paper, we propose a method that combines semantic segmentation information and spatial motion information of associated pixels to cope with dynamic objects based on ORB-SLAM2. We add a deep segmentation network SegNet to segment input image and obtain the semantic information for each feature point. Next, the spatial velocity of feature points between adjacent frames is calculated assuming uniform motion. Finally, the two parts are fused for the final judgment, and the dynamic feature points are removed to improve positioning accuracy. We evaluate our SLAM algorithms using the public KITTI dataset. The proposed algorithm has a similar overall accuracy level to ORB-SLAM2, but it is more accurate in sequences with many dynamic objects. On KITTI’s raw data sequence containing multiple dynamic objects, our pipeline achieves the best performance, improving 39.5% compared with the original ORB-SLAM2 system. We compare our algorithm with other state-of-the-art SLAM systems used to cope with dynamic environments. The results show that the proposed algorithm has better performance.

The integration of cloud computing and edge computing is an effective way to achieve global consistent and real-time multi-robot Simultaneous Localization and Mapping (SLAM). Cloud computing effectively solves the problem of limited computing, communication and storage capacity of terminal equipment. However, limited bandwidth and extremely long communication links between terminal devices and the cloud result in serious performance degradation of multi-robot SLAM systems. To reduce the computational cost of feature tracking and improve the real-time performance of the robot, a lightweight SLAM method of optical flow tracking based on pyramid IMU prediction is proposed. On this basis, a centralized multi-robot SLAM system based on a robot-edge-cloud layered architecture is proposed to realize real-time collaborative SLAM. It avoids the problems of limited on-board computing resources and low execution efficiency of single robot. In this framework, only the feature points and keyframe descriptors are transmitted and lossless encoding and compression are carried out to realize real-time remote information transmission with limited bandwidth resources. This design reduces the actual bandwidth occupied in the process of data transmission, and does not cause the loss of SLAM accuracy caused by data compression. Through experimental verification on the EuRoC dataset, compared with the current most advanced local feature compression method, our method can achieve lower data volume feature transmission, and compared with the current advanced centralized multi-robot SLAM scheme, it can achieve the same or better positioning accuracy under low computational load. Note to Practitioners —The purpose of this paper is to reduce the communication load of a Cloud-Edge-Robot system by compressing and transmitting of keyframes and non-keyframes, respectively, which is suitable for a multi-robot SLAM system and can realize multi-robot joint localization and sparse map reconstruction under efficient communication. Currently, remote SLAM or centralized multi-robot SLAM is usually implemented by transferring the whole image or the features and descriptors of the image. In this paper, lightweight SLAM optical flow tracking based on pyramid IMU prediction is implemented to track non-keyframes. At the edge server, tracking between non-keyframes is realized only by transmitting keypoints. For keyframes, the pose estimation is realized by transmitting compressed features and descriptors. Multi-robot localization and map fusion are realized in the cloud through key frame feature information. Experiments on public datasets show that this method is feasible and can achieve high-precision joint positioning with a low amount of transmitted data. In future studies, we will apply this framework to more real-world systems, while achieving rich, accurate map fusion with more advanced features.

The most commonly used simultaneous localization and mapping (SLAM) scheme often assumes a static environment, leading to significant errors in pose estimation when operating in highly dynamic scenes. To address this limitation and improve the robustness and accuracy of positioning in dynamic environments, this study proposes CD-SLAM, a real-time stereo vision inertial SLAM system specifically designed for complex dynamic environments, based on ORB-SLAM3. CD-SLAM enhances the tracking thread and introduces a new parallel thread that utilizes YOLOv5 to detect objects in each input frame and extract semantic information. This semantic information, combined with prior information from the inertial measurement unit (IMU), is used for pose estimation, eliminating the pose information of dynamic objects and consequently improving the accuracy and robustness of positioning. Furthermore, CD-SLAM employs scene flow to calculate the distance between adjacent frames and determine the spatial velocity between them, compensating for potential static information through a velocity filtering algorithm. To enhance positioning accuracy in challenging environments with weak textures, CD-SLAM integrates an IMU for motion prediction and coherence detection. Finally, appeal information is integrated to determine the motion status of objects in the scene and filter out dynamic feature points. Experimental tests conducted on the VIODE dataset demonstrate that CD-SLAM outperforms the existing algorithms in terms of accuracy and robustness.

Intelligent mobile vehicles asks for the ability to learn to comprehend immediate surroundings similar to human cognition and autonomously navigate in an unknown scene. Although simultaneous localization and mapping (SLAM) can construct a geometrically surrounding map, the map contains little semantic information. This paper presents a novel stereo visual-inertial system with loop closure detection based on a semantic-topological map framework. In our system, a hybrid 3D point cloud semantic-topological mapping framework is used to realize autonomous navigation simply by providing a map for path planning and meanwhile for storing semantic information under a dynamic environment. A stereo visual-inertial system can obtain more accurate visual-inertial odometry than a monocular system, and a visual-inertial system is capable of real-time map correction when the localization module detects a loop closure. This paper adopts state-of-the-art Mask R-CNN components to obtain 2D semantic information and project it to a semi-dense 3D semantic-topological map built by stereo visual-inertial SLAM. We perform experimental evaluation of our proposed system on the EuRoC MAV datasets and we also compare it with other competitive methods. Our results demonstrate that the proposed method is more effective than the other competitive methods.

Traditional visual-inertial SLAM (Simultaneous Localization and Mapping) systems predominantly rely on feature point matching from a single robot to realize the robot pose estimation and environment map construction. However, in complex scenarios, these traditional systems struggle with issues such as tracking failures due to illumination changes, rapid movements, and low-texture environments, and they perform poorly in terms of mapping efficiency and global consistency. To address these challenges, we propose a centralized collaborative SLAM system that employs both point and line features for tracking in the robot and map fusion in the cloud. The proposed system leverages the fusion of point and line features across all instances in the process, which allows our method to achieve higher localization accuracy in structured, low-texture scenes. With the aid of classifying gravity-aligned vertical lines and spatial parallel lines, the proposed system can deliver faster and more accurate odometry in complex scenes. Furthermore, we developed intra-robot and inter-robot loop closure detection methods based on point and line features, generating a globally consistent sparse point cloud and structured scene map in the cloud. Our method is able to build richer maps while improving accuracy compared to existing methods. Experimental results on public datasets and in real-world environments show that, compared to existing advanced methods, our approach demonstrates better performance.

Combining multiple sensors for environment sensing and self-positioning is significant for automatic driving. This paper proposes a novel simultaneous localization and mapping (SLAM) system framework that integrates the information ofmultiple sensors including camera, Light Detection and Ranging (LiDAR), Inertial Measurement Unit (IMU), and Global Positioning System (GPS) based on vision-LiDAR calibration. Different sensors are fused in a tightly coupled manner and finally optimized by a factor graph. The automatic vision-LiDAR calibration (AVLC) is proposed in this paper to reduce the error caused by the unexpected change in the sensor. Further, the semantic map is established by the target detection module, which provides convenience for navigation and obstacle avoidance. The proposed algorithm uses the Complex-YOLO for 3D object recognition and then combines the recognition results with the semi-dense point cloud map generated by the multi-sensor fusion positioning algorithm with AVLC to achieve the purpose of enriching map information. Extensive experiments on multiple datasets show that the proposed algorithm has higher accuracy and robustness than other existing algorithms.