Hyukdoo Choi

CV-SLAM using ceiling boundary

This paper deals with simultaneous localization and mapping(SLAM) problem for a mobile robot that travels around the indoor environments. A single camera looking up the ceiling is used as the only sensor. Line features are extracted from the boundaries between the ceiling and walls and parameterized for SLAM update. Extended Kalman Filter(EKF) is used for simultaneously estimating the current robot pose and building a map with the line features. When the robot is kidnapped, Monte Carlo Localization(MCL) is used for finding the robot pose. To improve the localization performance, the resampling method is modified. The experiment is practiced in our indoor test bed and the proposed algorithms are proved by the experimental results.

Simultaneous Global Localization and Mapping

This paper proposes a hybrid approach to global localization and simultaneous localization and mapping (SLAM). Global localization and SLAM techniques have been independently developed but now researchers seek to simultaneously solve two problems regarding localization with an imperfect map and no a priori state information. Until now, integration of global localization and SLAM have not undergone extensive research. We propose a new approach for the new problem, called simultaneous global localization and mapping (SiGLAM). Our method is derived from the feature-driven method of global localization but evolved to be more robust to sensor noise and imperfections in a map. We do not wait until the only hypothesis survives. Hypotheses are continuously generated and managed in a conservative way. Instead, the best hypothesis is selected by hypothesis scoring. We demonstrate the proposed algorithm with simulations and real-world experiments. The results prove that our method outperforms other existing methods.

Grid mapping adaptive to various map sizes for Sbot

Grid mapping is a fundamental mapping algorithm in the mobile robotics. This paper proposes an indoor grid mapping technique which is adaptive to various map size. Rao-Blackwellized Particle Filters (RBPF) have been used to build a map. Each mapping procedures has the different number of particles. Small area uses more particles than large area. The suitable number of particles can improve the map quality and leads to low computation cost. Experiments have been conducted with unmanned delivering robot named Sbot. It is equipped with laser scanners and sonar sensors. We demonstrated the mapping system in different indoor environments and proved its efficiency.

An Efficient Ceiling-view SLAM Using Relational Constraints Between Landmarks

In this paper, we present a new indoor simultaneous localization and mapping‘ (SLAM) technique based on an upward-looking ceiling camera. Adapted from our previous work [17], the proposed method employs sparsely-distributed line and point landmarks in an indoor environment to aid with data association and reduce extended Kalman filter computation as compared with earlier techniques. Further, the proposed method exploits geometric relationships between the two types of landmarks to provide added information about the environment. This geometric information is measured with an upward-looking ceiling camera and is used as a constraint in Kalman filtering. The performance of the proposed ceiling-view (CV) SLAM is demonstrated through simulations and experiments. The proposed method performs localization and mapping more accurately than those methods that use the two types of landmarks without taking into account their relative geometries.

A new cvSLAM exploiting a partially known landmark association

There are some situations in which the landmarks used in simultaneous localization and mapping (SLAM) have their own classes and for ceiling view (CV)-based navigation, this is usually the case. Ceilings in the home or the office have circular landmarks, such as lamps, speakers, fire alarms, smoke alarms, and so on, but to our knowledge, their classes have not been fully exploited in the data association of SLAM. In this paper, a new SLAM method that exploits the class of the landmarks is proposed and is applied to ceiling view-based SLAM (cvSLAM). The fact that the landmark classification cannot always be correct is also taken into account in the new SLAM and is formulated in the FastSLAM framework. Finally, simulations and experiments are conducted and the validity of the proposed method is demonstrated.

Scan likelihood evaluation in FastSLAM using binary Bayes filter

FastSLAM is a fundamental algorithm for Simultaneous Localization and Mapping (SLAM). FastSLAM based on grid map is a popular method to build a map of both the structured and unstructured environment. The performance of FastSLAM significantly depends on evaluation of measurement likelihood. In this paper, we propose a new method to evaluate laser scan likelihood using the binary Bayes filter. This method supports the right particles but does not suffer from particle depletion problem. We implemented the hardware system based on the Pioneer 2-DX platform equipped with the Hokuyo laser scanner. The experimental result shows that the proposed method builds the map accurately.

Mobile robot tracking in the BSN using support vector machine

This paper presents a localization algorithm for mobile node, such as robot in the binary sensor network. The binary sensor network can be constructed with any systems, wireless LANs or other communication methods. Our proposed location estimation process is based on support vector machine. We find optimal separating hyperplane which classifies sensor groups, and we estimate the position of mobile robot with radio range constraints. The proposed method can be used in the noisy communication environment. Simulation results are given to show the better performance of the proposed method than other methods.

Map management system for cv-SLAM

This paper present a map management system for ceiling vision (cv)-SLAM where a map is comprised of line landmarks on the ceiling. Since the size of a map, which is the number of landmarks, is related to computational and memorial cost, it should be managed in an appropriate size. For this purpose, the binary Bayes filter is utilized to estimate the probability of existence of a landmark. We demonstrate the system in the real world experiment to prove its performance.

Mobile Robot Exploration via Pseudo Range Model

This paper presents an autonomous exploration system using a robot equipped with a vision sensor but no range sensor. The fully automatic exploration system operating on a mobile robot determines the next motion in the environment. We extract ceiling boundaries from images. Then we use the pseudo range model that converts the ceiling boundary to range information. This paper proposes the vision based exploration system and presents experimental results.

GraphSLAM Improved by Removing Measurement Outliers

This paper presents the GraphSLAM improved by selecting the measurement with respect to their likelihoods. GraphSLAM estimates the robot`s path and map by utilizing the entire history of input data. However, GraphSLAM`s performance suffers a lot from severely noisy measurements. In this paper, we present GraphSLAM improved by the selective measurement method. Thus the presented GraphSLAM provides higher performance compared with the standard GraphSLAM.