Visual inertial odometry and SLAM algorithms are widely used in various fields, such as service robots, drones, and autonomous vehicles. Most of the SLAM algorithms are based on assumption that landmarks are static. However, in the real-world, various dynamic objects exist, and they degrade the pose estimation accuracy. In addition, temporarily static objects, which are static during observation but move when they are out of sight, trigger false positive loop closings. To overcome these problems, we propose a novel visual-inertial SLAM framework, called DynaVINS , which is robust against both dynamic objects and temporarily static objects. In our framework, we first present a robust bundle adjustment that could reject the features from dynamic objects by leveraging pose priors estimated by the IMU preintegration. Then, a keyframe grouping and a multi-hypothesis-based constraints grouping methods are proposed to reduce the effect of temporarily static objects in the loop closing. Subsequently, we evaluated our method in a public dataset that contains numerous dynamic objects. Finally, the experimental results corroborate that our DynaVINS has promising performance compared with other state-of-the-art methods by successfully rejecting the effect of dynamic and temporarily static objects.

With the increasing demand for autonomous systems in the field of inspection, the use of unmanned aerial vehicles (UAVs) to replace human labor is becoming more frequent. However, the Global Positioning System (GPS) signal is usually denied in environments near or under bridges, which makes the manual operation of a UAV difficult and unreliable in these areas. This paper addresses a novel hierarchical graph-based simultaneous localization and mapping (SLAM) method for fully autonomous bridge inspection using an aerial vehicle, as well as a technical method for UAV control for actually conducting bridge inspections. Due to the harsh environment involved and the corresponding limitations on GPS usage, a graph-based SLAM approach using a tilted 3D LiDAR (Light Detection and Ranging) and a monocular camera to localize the UAV and map the target bridge is proposed. Each visual-inertial state estimate and the corresponding LiDAR sweep are combined into a single subnode. These subnodes make up a “supernode” that consists of state estimations and accumulated scan data for robust and stable node generation in graph SLAM. The constraints are generated from LiDAR data using the normal distribution transform (NDT) and generalized iterative closest point (G-ICP) matching. The feasibility of the proposed method was verified on two different types of bridges: on the ground and offshore.

Localization has been a challenging task for autonomous navigation. A loop detection algorithm must overcome environmental changes for the place recognition and re-localization of robots. Therefore, deep learning has been extensively studied for the consistent transformation of measurements into localization descriptors. Street view images are easily accessible; however, images are vulnerable to appearance changes. LiDAR can robustly provide precise structural information. However, constructing a point cloud database is expensive, and point clouds exist only in limited places. Different from previous works that train networks to produce shared embedding directly between the 2D image and 3D point cloud, we transform both data into 2.5D depth images for matching. In this work, we propose a novel cross-matching method, called (LC)$^{2}$ , for achieving LiDAR localization without a prior point cloud map. To this end, LiDAR measurements are expressed in the form of range images before matching them to reduce the modality discrepancy. Subsequently, the network is trained to extract localization descriptors from disparity and range images. Next, the best matches are employed as a loop factor in a pose graph. Using public datasets that include multiple sessions in significantly different lighting conditions, we demonstrated that LiDAR-based navigation systems could be optimized from image databases and vice versa.

In this paper, we propose a generalized grouping and pruning method for RGB-D SLAM in low-dynamic environments. The conventional grouping and pruning methods successfully reject the effect of dynamic objects in pose graph optimization (PGO). However, these methods sometimes fail when high-dynamic objects are dominant in the images captured by RGB-D sensors. Furthermore, once it is determined whether the features from dynamic objects are included in some nodes, the corresponding nodes are entirely removed even though these nodes partially include true constraints, which leads to an inaccurate PGO. To tackle these problems, we propose a novel method with intra-grouping, inter-grouping, and selective pruning, called G2P-SLAM. Accordingly, our method successfully rejects false constraints from dynamic objects selectively, thus preserving true constraints from static objects as many as possible. As experimentally verified on both our own datasets and public datasets, our proposed method shows promising performance compared with the state-of-the-art methods. Furthermore, experimental results corroborate that our G2P-SLAM enables robust PGO in both dynamic and low-dynamic environments.

Despite extensive research in robust visual-inertial navigation systems (VINS) in dynamic environments, many approaches remain vulnerable to objects that suddenly start moving, which are referred to as abruptly dynamic objects. In addition, most approaches have considered the effect of dynamic objects only at the feature association level. In this study, we observed that the state estimation diverges when errors from false correspondences owing to moving objects incorrectly propagate into the IMU bias terms. To overcome these problems, we propose a robust VINS framework called DynaVINS++, which employs a) adaptive truncated least square method that adaptively adjusts the truncation range using both feature association and IMU preintegration to effectively minimize the effect of the dynamic objects while reducing the computational cost, and b) stable state recovery with bias consistency check to correct misestimated IMU bias and to prevent the divergence caused by abruptly dynamic objects. As verified in both public and real-world datasets, our approach shows promising performance in dynamic environments, including scenes with abruptly dynamic objects.

In this paper, a novel wall-climbing drone installed with a rotary arm is proposed for climb operation on various shaped walls. Robots that can climb the wall are applicable to many operations such as structural health monitoring of high-rise buildings, bridges, nuclear power plants; maintenance of solar panels and glass walls; and visual inspection of the aircraft and vessels. Traditional wall-climbing robots use magnetic, vacuum, or bioinspired methods mimicking gecko foot for those purposes. However, those methods cannot be applied to various-shaped walls like inclined, and an obstacle existing walls such as window frames. To solve this problem, unmanned aerial vehicle (UAV)-type wall-climbing robots have recently been developed, but those types have disadvantages in terms of energy efficiency and impact in attaching and detaching processes. Therefore, in this paper, a wall-climbing drone with a rotary arm is proposed. The angle of the rotary arm is controlled depending on the slope angle of the wall to achieve energy efficiency. The drone was prototyped and tested in indoor environment to verify the feasibility of the mechanism.