Taekjun Oh

Magnetic field constraints and sequence-based matching for indoor pose graph SLAM

The objective of pose graph optimization is to estimate the robot trajectory from the constraints of relative pose measurements. Since the magnetic field in indoor environments is stable in the temporal domain and sufficiently varying in the spatial domain, we can exploit these characteristics to generate the constraints of the pose graph. In this paper we provide a method of solving a simultaneous localization and mapping (SLAM) problem by employing pose graph optimization and indoor magnetic measurements. Specifically, different types of constraints for local heading correction and global loop closing, respectively, are designed. For the loop closing constraints in particular, we first examine spatial similarity of the indoor magnetic field and verify that the use of measurement sequences rather than a single measurement mitigates the ambiguity of the magnetic measurements. A loop closing algorithm is then proposed based on the sequence of magnetic measurement and applied to the pose graph optimization. Experimental results show that the proposed SLAM system with only wheel encoders and a single magnetometer obtains comparable results with a reference-level SLAM system in terms of robot trajectory, thereby validating the feasibility of applying magnetic constraints to indoor pose graph SLAM. We design pose graph constraints using indoor magnetic field measurements.Magnetic measurements locally aid in the correction of heading directions.Use of a sequence of magnetic measurements aids in the detection of loop closures.Performance of a system with a single magnetometer and wheel encoders is evaluated.

Graph Structure-Based Simultaneous Localization and Mapping Using a Hybrid Method of 2D Laser Scan and Monocular Camera Image in Environments with Laser Scan Ambiguity

Localization is an essential issue for robot navigation, allowing the robot to perform tasks autonomously. However, in environments with laser scan ambiguity, such as long corridors, the conventional SLAM (simultaneous localization and mapping) algorithms exploiting a laser scanner may not estimate the robot pose robustly. To resolve this problem, we propose a novel localization approach based on a hybrid method incorporating a 2D laser scanner and a monocular camera in the framework of a graph structure-based SLAM. 3D coordinates of image feature points are acquired through the hybrid method, with the assumption that the wall is normal to the ground and vertically flat. However, this assumption can be relieved, because the subsequent feature matching process rejects the outliers on an inclined or non-flat wall. Through graph optimization with constraints generated by the hybrid method, the final robot pose is estimated. To verify the effectiveness of the proposed method, real experiments were conducted in an indoor environment with a long corridor. The experimental results were compared with those of the conventional GMappingapproach. The results demonstrate that it is possible to localize the robot in environments with laser scan ambiguity in real time, and the performance of the proposed method is superior to that of the conventional approach.

Mobile robot localization by matching 2D image features to 3D point cloud

In this paper we describe a method for solving a mobile robot localization problem using prior data. By matching 2D image features to a 3D point cloud, the robot position is estimated in the prior point cloud. We prove our method by testing at specific locations over the whole point clod data.

A Probabilistic Feature Map-Based Localization System Using a Monocular Camera

Image-based localization is one of the most widely researched localization techniques in the robotics and computer vision communities. As enormous image data sets are provided through the Internet, many studies on estimating a location with a pre-built image-based 3D map have been conducted. Most research groups use numerous image data sets that contain sufficient features. In contrast, this paper focuses on image-based localization in the case of insufficient images and features. A more accurate localization method is proposed based on a probabilistic map using 3D-to-2D matching correspondences between a map and a query image. The probabilistic feature map is generated in advance by probabilistic modeling of the sensor system as well as the uncertainties of camera poses. Using the conventional PnP algorithm, an initial camera pose is estimated on the probabilistic feature map. The proposed algorithm is optimized from the initial pose by minimizing Mahalanobis distance errors between features from the query image and the map to improve accuracy. To verify that the localization accuracy is improved, the proposed algorithm is compared with the conventional algorithm in a simulation and realenvironments.

Feature-Based 6-DoF Camera Localization Using Prior Point Cloud and Images

In this paper, we present a new localization algorithm to estimate the localization of a robot based on prior data. Over the past decade, the emergence of numerous ways to utilize various prior data has opened up possibilities for their applications in robotics technologies. However, challenges still remain in estimating a robot’s 6-DoF position by simply analyzing the limited information provided by images from a robot. This paper describes a method of overcoming this technical hurdle by calculating the robot’s 6-DoF location. It only utilizes a current 2D image and prior data, which consists of its corresponding 3D point cloud and images, to calculate the 6-DoF position. Furthermore, we employed the SURF algorithm to find the robot’s position by using the image’s features and the 3D projection method. Experiments were conducted by the loop of 510m trajectory, which is included the prior data. It is expected that our method can be applied to broad areas by using enormous data such as point clouds and street views in the near future.

Image-based localization using prior map database and Monte Carlo Localization

The aim of this paper is to propose an image and map data-based localization method applicable to a variety of environments. For the localization, we use prior map database, image-based localization method, and MCL (Monte Carlo Localization). The results were confirmed by open data set in a variety of environments. The experimental results show the feasibility of the proposed method for the robot localization.

Indoor Magnetic Pose Graph SLAM with Robust Back-End

In this paper, a method of solving a simultaneous localization and mapping (SLAM) problem is proposed by employing pose graph optimization and indoor magnetic field measurements. The objective of pose graph optimization is to estimate the robot trajectory from the constraints of relative pose measurements. Since the magnetic field in indoor environments is stable in a temporal domain and sufficiently varying in a spatial domain, these characteristics can be exploited to generate the constraints in pose graphs. In this paper two types of constraints are designed, one is for local heading correction and the other for loop closing. For the loop closing constraint, sequence-based matching is employed rather than a single measurement-based one to mitigate the ambiguity of magnetic measurements. To improve the loop closure detection we further employed existing robust back-end methods proposed by other researchers. Experimental results show that the proposed SLAM system with only wheel encoders and a single magnetometer offers comparable results with a reference-level SLAM system in terms of robot trajectory, thereby validating the feasibility of applying magnetic constraints to the indoor pose graph SLAM.

Accurate Localization in Urban Environments Using Fault Detection of GPS and Multi-sensor Fusion

In order to make robots perform tasks autonomously, it is necessary for robots to know the surrounding environments. Therefore, a world modeling should be made in advance or concurrently. It is important to know an accurate position for the accurate world modeling. The aim of this paper is an accurate localization method for the world modeling under the situation where the portion of signals from global positioning system (GPS) satellites is blocked in urban environments. In this paper, we propose a detection method for non-line-of-sight satellites and a localization method using the GPS, the inertial measurement unit (IMU), the wheel encoder, and the laser range finder (LRF). To decide whether the signal from the satellite is blocked by the building, the local map that is made from the local sensors and an LRF is exploited. Then the GPS reliability is established adaptively in a non-line-of-sight situation. Through an extended Kalman filter (EKF) with the GPS reliability the final robot pose is estimated. To evaluate the performance of the proposed methods, the accuracy of the proposed method is analyzed using ground truth from Google maps. Experimental results demonstrate that the proposed method is suitable for the urban environments.

Graph-based SLAM approach for environments with laser scan ambiguity

The study of UGVs (Unmanned Ground Vehicles) has been actively promoted and researched. In order to operate unmanned vehicles or robots autonomously, it is necessary to understand the surrounding environment and know where it is. This paper proposes a novel localization method using a monocular camera and a laser scanner for a robot in an environment that is difficult to localize. We exploited the hybrid method between depth data from a laser scanner and the image feature of the camera, and then the collected robot pose information is estimated using the pose graph structure. In order to verify the performance of the proposed algorithm, experiments are conducted in an indoor environment. The results of the proposed algorithm are then numerically compared with the results of the conventional method. We confirmed that it is possible to localize the robot in environments with laser scan ambiguity such as a long corridor.

Image-based localization using image database and local 3D maps

Image-based localization is one of the most important techniques for autonomous navigation. This paper proposes a 6-DoF localization method based on image database and local 3D maps. To perform localization, the image database and local 3D map of each image are collected in advance. By applying a fast place recognition algorithm such as FAB-MAP, an initial position is assumed to be available at a past position of a matched image from the database. Then, a 6-DoF camera pose is estimated using PnP algorithm by 2D-to-3D matching correspondences in a query image and a local 3D map. The proposed algorithm is demonstrated in real environment.