Heon-Cheol Lee

Improved particle fusing geometric relation between particles in fastslam

FastSLAM is a framework for simultaneous localization and mapping using a Rao-Blackwellized particle filter (RBPF). But, FastSLAM is known to degenerate over time due to the loss of particle diversity, mainly caused by the particle depletion problem in resampling phase. In this work, improved particle filter using geometric relation between particles is proposed to restrain particle depletion and to reduce estimation errors and error variances. It uses a KD tree (k-dimensional tree) to derive geometric relation among particles and filters particles with importance weight conditions for resampling. Compared to the original particle filter used in FastSLAM, this technique showed less estimation error with lower error standard deviation in computer simulations.

Probabilistic map merging for multi-robot rbpf-slam with unknown initial poses

This paper addresses the map merging problem, which is the most important issue in multi-robot simultaneous localization and mapping (SLAM) using the Rao-Blackwellized particle filter (RBPF-SLAM) with unknown initial poses. The map merging is performed using the map transformation matrix and the pair of map merging bases (MMBs) of the robots. However, it is difficult to find appropriate MMBs because each robot pose is estimated under multi-hypothesis in the RBPF-SLAM. In this paper, probabilistic map merging (PMM) using the Gaussian process is proposed to solve the problem. The performance of PMM was verified by reducing errors in the merged map with computer simulations and real experiments.

PSO-FastSLAM: An improved FastSLAM framework using particle swarm optimization

FastSLAM is a framework which solves the problem of simultaneous localization and mapping using a Rao-Blackwellized particle filter. Conventional FastSLAM is known to degenerate over time in terms of accuracy due to the particle depletion in resampling phase. In this work, a new FastSLAM framework is proposed to prevent the degeneracy by particle cooperation. First, after resampling phase, a target that represents an estimated robot position is computed using the positions of particles. Then, particle swarm optimization is performed to update the robot position by means of particle cooperation. Computer simulations revealed that the proposed framework shows lower RMS error in both robot and feature positions than conventional FastSLAM.

Analysis of Resampling Process for the Particle Depletion Problem in FastSLAM

The state-of-the-art FastSLAM has been shown to cause a particle depletion problem while performing simultaneous localization and mapping for mobile robots. As a result, it always produces over-confident estimates of uncertainty as time progresses. This particle depletion problem is mainly due to the resampling process in FastSLAM, which tends to eliminate particles with low weights. Therefore, the number of particles to perform loop-closure decreases, which makes the performance of FastSLAM degenerate. The resampling process has not been thoroughly analyzed even though it is the main reason for the particle depletion problem. In this paper, standard resampling algorithms (systematic residual and partial resampling), and a rank-based resampling applying genetic algorithms are analyzed using computer simulations. For the thorough analysis, several performance measures are used such as effective sample size, the number of distinct particles, root mean square (RMS) errors, and complexity. According to the simulation results, all resampling algorithms could not resolve the particle depletion problem.

Adaptive prior boosting technique for the efficient sample size in fastSLAM

FastSLAM has been shown to degenerate over time due to sample impoverishment, that is, poor samples are generated during the sampling process. One of major culprits of the sample impoverishment problem is lack of the number of particles estimating the pose of the robot and the environment. In this work, an adaptive prior boosting technique is proposed for the efficient sample size according to the uncertainty of each situation in performing FastSLAM. It uses a back-propagation neural network, learned in various environments, in order to decide the required sample size. This adaptive approach generates a small number of particles when the uncertainty is low while performing FastSLAM, and it generates a large number of particles when the uncertainty is high. This technique efficiently generates the sample size in computer simulations and real environmental experiments, which significantly reduces the RMS feature and position errors.

Improved Feature Map Merging Using Virtual Supporting Lines for Multi-Robot Systems

This paper addresses the problem of feature map merging, which is one of the essential techniques for multi-robot systems. If inter-robot measurements are not available for feature map merging, the only way to obtain the map transformation matrix is feature map matching. However, the conventional feature map matching technique requires too much computation time because it has to be iteratively performed to compute the degree of the mismatch between multiple feature maps. This paper proposes a non-iterative feature map merging technique using virtual supporting lines (VSLs) which is also accurate and robust. The proposed technique extracts the spectral information of multiple feature maps using VSLs and obtains the map transformation matrix using the circular cross-correlation between the extracted spectral information of the multiple feature maps. The proposed technique was tested on feature maps produced by experiments with vision sensors, which was performed non-iteratively. In addition, it consistently showed a high acceptance index, which indicates the degree of accuracy for feature map merging.

A survey of map merging techniques for cooperative-SLAM

This paper presents a survey of map merging techniques for cooperative-SLAM. The recently proposed map merging techniques are classified into two categories: direct map merging and indirect map merging. In each category, several techniques are briefly described. Then, their advantages and disadvantages are discussed in the context of accuracy and computation time. The description and discussion can contribute to realizing and improving cooperative-SLAM.

Grafting: A Path Replanning Technique for Rapidly-Exploring Random Trees in Dynamic Environments

Abstract The rapidly-exploring random trees (RRT) is a sampling-based path planner which utilizes simultaneously kinematics and dynamics of a robot. However, since the RRT has produced a robot path without taking the existence of dynamic obstacles into consideration, RRT-based navigation has the risk of a collision with dynamic obstacles. We proposed a path replanning technique for the RRT applied to robot navigation in dynamic environments, which is named grafting. The proposed technique replans a safe and efficient path in real time instead of the original path which may cause a collision with dynamic obstacles. Moreover, the replanned path can be easily merged into the original RRT path because the grafting technique preserves the property of the RRT. The grafting technique was tested by simulations in various dynamic environments, which revealed that the grafting technique was capable of replanning a safe and efficient path for RRT-based navigation in real time.

Comparison and analysis of scan matching techniques for Cooperative-SLAM

A scan matching technique is a key technique to implement Cooperative-SLAM (C-SLAM). Although a variety of scan matching techniques have been developed, each of them has both merits and faults. This paper compares and analyzes three well-known scan matching techniques which are ICP (Iterative Closest Points), HSM (Hough Scan Matching), and PSM (Polar Scan Matching). The comparison and analysis are carried out in terms of error and computation time for scan pairs which consists of a reference scan and a current scan. Moreover, the scan matching technique which shows the best performance is applied to C-SLAM.

Robust scan matching with curvature-based matching region selection

This paper presents a novel scan matching algorithm which uses not whole scan region but only salient scan region selected around curvature-based features. A curvature function computed by the relative coordinates of neighbor scan points is used to extract salient features which are invariant to translation and rotation. Because the scan matching regions are selected around the salient features, the presented algorithm can be robustly performed even in noisy environments. The robustness of the presented algorithm was tested by datasets obtained from various noisy environments and was verified by consistently showing smaller errors than other scan matching algorithms. Moreover, the presented algorithm was successfully applied to SLAM.