This letter presents Switch-SLAM, switching-based LiDAR-inertial-visual SLAM for degenerate environments, designed to tackle the challenges in degenerate environments for LiDAR and visual SLAM. Switch-SLAM achieves high robustness and accuracy by utilizing a switching structure that transitions from LiDAR to visual odometry when degeneration of LiDAR odometry is detected. To efficiently detect degeneration, Switch-SLAM incorporates a non-heuristic degeneracy detection method that does not require heuristic tuning and demonstrates generalizability across various environments. Switch-SLAM is evaluated on diverse datasets containing both LiDAR and visual odometry degeneracy scenarios. The experimental results highlight the accurate and robust localization by the proposed method in multiple challenging environments with either LiDAR or visual SLAM degeneracy.

In this paper, we propose a path planning method for the localization of radiation sources using a mobile robot equipped with an imaging gamma-ray detector, which has a field of view in all directions. The ability to detect and localize radiation sources is essential for ensuring nuclear safety, security, and surveillance. To enable the autonomous localization of radiation sources, the robot must have the ability to automatically determine the next location for gamma ray measurement instead of following a predefined path. The number of incident events is approximated to be the squared inverse proportional to the distance between the radiation source and the detector. Therefore, the closer the distance to the source, the shorter the time required to obtain the same radiation counts measured by the detector. Hence, the proposed method is designed to reduce this distance to a position where a sufficient number of gamma-ray events can be obtained; then, a path to surround the radiation sources is generated. The proposed method generates this path by performing principal component analysis based on the results obtained from previous measurements. Both simulations and actual experiments demonstrate that the proposed method can automatically generate a measurement path and accurately localize radiation sources.

At modern construction sites, utilizing GNSS (Global Navigation Satellite System) to measure the real-time location and orientation (i.e. pose) of construction machines and navigate them is very common. However, GNSS is not always available. Replacing GNSS with on-board cameras and visual simultaneous localization and mapping (visual SLAM) to navigate the machines is a cost-effective solution. Nevertheless, at construction sites, multiple construction machines will usually work together and side-by-side, causing large dynamic occlusions in the cameras' view. Standard visual SLAM cannot handle large dynamic occlusions well. In this work, we propose a motion segmentation method to efficiently extract static parts from crowded dynamic scenes to enable robust tracking of camera ego-motion. Our method utilizes semantic information combined with object-level geometric constraints to quickly detect the static parts of the scene. Then, we perform a two-step coarse-to-fine ego-motion tracking with reference to the static parts. This leads to a novel dynamic visual SLAM formation. We test our proposals through a real implementation based on ORB-SLAM2, and datasets we collected from real construction sites. The results show that when standard visual SLAM fails, our method can still retain accurate camera ego-motion tracking in real-time. Comparing to state-of-the-art dynamic visual SLAM methods, ours shows outstanding efficiency and competitive result trajectory accuracy. GRAPHICAL ABSTRACT

In robot teleoperation, a lack of depth information often results in collisions between the robots and obstacles in its path or surroundings. To address this issue, free viewpoint images can greatly benefit the operators in terms of collision avoidance as the operators are able to view the robot’s surrounding from the images at arbitrary points, giving them a better depth information. In this paper, a novel free viewpoint image generation system is proposed. One approach to generate free viewpoint images is to use multiple cameras and Light Detection and Ranging (LiDAR). Instead of using the expensive LiDAR, this study utilizes a cost-effective laser rangefinder (LRF) and a characteristic of man-made environments. In other words, we install multiple fisheye cameras and an LRF on a robot. Free viewpoint images are generated under the assumption that walls are perpendicular to the floor. Furthermore, an easy calibration for estimating the poses of the multiple fisheye cameras, the LRF, and the robot model is proposed. Experimental results show that the proposed method can generate free viewpoint images using cameras and an LRF. Finally, the proposed method is primarily implemented using OpenGL Shading Language to utilize a graphics processing unit computation to achieve a real-time processing of the multiple high-resolution images. Supplementary videos and our source code are available at our project page (https://matsuren.github.io/fvp).

In this study, we developed a novel method to determine the optimal viewpoint from which an operator could realize faster and more accurate robot teleoperation using reinforcement learning. The reinforcement learning model was trained using images obtained from several candidate viewpoints from scratch, and the viewpoint at which the model achieved the highest rewards was considered the optimal viewpoint. The experimental results validated the choice of the world model as the reinforcement learning model. Moreover, subject experiments wherein subjects operated a robot arm remotely to reach the target position were conducted. The experiments produced results that strongly aligned with the performance obtained through computer simulations, indicating that the proposed method is capable of selecting the optimal viewpoint without handcrafted design and subject experiments.

In this paper, we propose the octave deep plane-sweeping network (OctDPSNet). OctDPSNet is a novel learning-based plane-sweeping stereo, which drastically reduces the required GPU memory and computation time while achieving a state-of-the-art depth estimation accuracy. Inspired by octave convolution, we divide image features into high and low spatial frequency features, and two cost volumes are generated from these using our proposed plane-sweeping module. To reduce spatial redundancy, the resolution of the cost volume from the low spatial frequency features is set to half that of the high spatial frequency features, which enables the memory consumption and computational cost to be reduced. After refinement, the two cost volumes are integrated into a final cost volume through our proposed pixel-wise “squeeze-and-excitation” based attention mechanism, and the depth maps are estimated from the final cost volume. We evaluate the proposed model on five datasets: SUN3D, RGB-D SLAM, MVS, Scenes11, and ETH3D. Our model outperforms previous methods on five datasets while drastically reducing the memory consumption and computational cost. Our source code is available at.