Point cloud is a discrete and unordered expression of 3D data. A lot of methods have been proposed to solve the problem in 3D object classification and scene recognition. To handle the huge amount of unordered point cloud, down-sampling before processing is needed. The shortage of existing sampling methods is the lack of geometry information consideration, which is essential for point cloud classification and segmentation tasks. Our method is mainly motivated by the observation that points with a high curvature variation can depict the outlines of objects. Thus, we propose a curvature variation based sampling method for point cloud classification and segmentation tasks. We aim to sample points with high curvature variations, which are considered to be more suitable for classification and segmentation tasks than the traditional sampling method. We combine the proposed sampling algorithm with the existing sampling method for multiple information fusion, and a higher accuracy and mean IoU can be achieved. The experimental results verify the advantage of considering curvature variation in classification and segmentation tasks.

Summary: Estimating the robot state within a known map is an essential problem for mobile robot; it is also referred to “localization”. Even LiDAR-based localization is practical in many applications, it is difficult to achieve global localization with LiDAR only for its low-dimension feedback, especially in environments with repetitive geometric features. A sensor-fusion-based localization system is introduced in this paper, which has the capability of addressing the global localization problem. Both LiDAR and vision sensors are integrated, making use of the rich information introduced by vision sensor and the robustness from LiDAR. A hybrid grid-map is built for global localization, and a visual global descriptor is applied to speed up the localization convergence, combined with a pose refining pipeline for improving the localization accuracy. Also, a trigger mechanism is introduced to solve kidnapped problem and verify the relocalization result. The experiments under different conditions are designed to evaluate the performance of the proposed approach, as well as a comparison with the existing localization systems. According to the experimental results, our system is able to solve the global localization problem, and the sensor-fusion mechanism in our system has an improved performance.

In the application of visual navigation, submap-based visual simultaneous localization and mapping (VSLAM) has become one of the most robust monocular solutions in recent years, which is able to resume tracking by multisubmap maintenance and merging. However, due to the lack of long-term data association between submaps, global consistency cannot be guaranteed in the existing work, especially in situations without loop-closure. Considering the fact that not all the submap have to be built in realtime, we propose a VSLAM system with realtime and nonrealtime hybrid style, RUMI-SLAM. Inspired by the rumination of mammalians that processes food in various stomaches and absorbs it in one stomach, RUMI-SLAM performs asynchronous submap building and centralized submap management. Building additional submaps in parallel leads to enriched mapping elements and enhanced data association across submaps. The experimental results demonstrate the superiority of RUMI-SLAM over the existing VSLAM systems, especially the robustness to challenging situations. We also provide real-robot experiments to demonstrate our RUMI-SLAM in the application of visual navigation. Our study provides a novel asynchronous submap-based VSLAM framework, which achieves globally consistent submap merging without the requirement of loop-closure.

Robustness in on-road driving Visual Odometry (VO) systems is critical, as it determines the reliable performance in various scenarios and environments. Especially with the development of data-driven technology, the combination of data-driven VO and model-based VO has achieved accurate tracking performance. However, the lack of generalization of pre-trained deep neural networks (DNN) limits the robustness of such a combination in unseen environments. In this study, we introduce a novel framework with appropriate usage of DNN prediction and improve the robustness in the self-driving application. Based on the characteristic of on-road self-driving motion and the DNN output, we propose a two-step optimization strategy with a variable degree of freedom (DoF), i.e., the use of two types of DoF representations during pose estimation. Specifically, our two-step optimization operates according to the residual of the optimization with the motion label classification from the pre-trained DNN, as well as our proposed Motion Evaluation by essential matrix construction. Experimental results show that our framework obtains better tracking accuracy than the existing methods.

Biped climbing robots have the character of multi-source asynchronous sensing, and they are powerful to replace humans to complete high-altitude and high-risk operations. As the key to autonomous climbing and visual perception, the VSLAM (Visual Simultaneous Localization And Mapping) for biped climbing robots is studied in this paper. Due to the limited computing power of the biped climbing robot, a cloud-based VSLAM system is introduced to offload the high computation. Considering the limited transmission in the high-altitude climbing environment, we propose a submap-based cloud-edge collaboration mechanism, in which a multiple submap VSLAM system is built. To satisfy the multiple source sensing of the biped climbing robot, we design an asynchronous submap building framework. According to the character of VSLAM, the edge-cloud transmission is optimized to reduce the bandwidth. The evaluations have verified our advanced performance in terms of bandwidth cost and tracking precision. Also, the experiment on the real climbing robot demonstrates our feasibility and superiority. Our Cloud-based Asynchronous Multi-submap VSLAM (CAMs-SLAM) system enables biped climbing robots to environment perception and achieves mapping and localization for later autonomous navigation.

Applying VSLAM to mobile robots with limited computing power is the key to achieving autonomous navigation, and the cloud-edge collaborative VSLAM is a solution. However, the VSLAM data transmission in the working site with limited communication is still an open problem. To solve this problem, two core issues should be considered: the transmission frequency and the transmission data volume. In this paper, we propose an asynchronous submap building framework to reduce the transmission frequency. Also, we design an implicit representationbased transmission method to save the transmission data volume while satisfying the data association between the edge and the cloud. Through the experiments, our method shows advanced performance in terms of communication demand with low transmission frequency and small data volume. At the same time, comparable precision to the state-of-the-art is achieved, showing the effectiveness of the data association building. Thanks to the reduced transmission frequency and data volume, our system provides a feasible way to the cloud-based VSLAM under limited communication.