Evaluation of the depth camera based SLAM algorithms

Similar documents
Design of a Pedestrian Detection System Based on OpenCV. Ning Xu and Yong Ren*

HIGH RESOLUTION DEPTH IMAGE RECOVERY ALGORITHM USING GRAYSCALE IMAGE.

#19 MONITORING AND PREDICTING PEDESTRIAN BEHAVIOR USING TRAFFIC CAMERAS

Pedestrian Protection System for ADAS using ARM 9

Active Pedestrian Safety: from Research to Reality

A Distributed Control System using CAN bus for an AUV

Pose Estimation for Robotic Soccer Players

Early Skip Decision based on Merge Index of SKIP for HEVC Encoding

Self-Driving Vehicles That (Fore) See

Deformable Convolutional Networks

Neural Network in Computer Vision for RoboCup Middle Size League

beestanbul RoboCup 3D Simulation League Team Description Paper 2012

Kinect-based Badminton Motion Sensor as Potential Aid for Coaching Strokes in Novice Level

Prediction of Basketball Free Throw Shooting by OpenPose

Research Article Research on Path Planning Method of Coal Mine Robot to Avoid Obstacle in Gas Distribution Area

CSU_Yunlu 2D Soccer Simulation Team Description Paper 2015

An approach for optimising railway traffic flow on high speed lines with differing signalling systems

Decompression Method For Massive Compressed Files In Mobile Rich Media Applications

The Design of Electrical Putter Car Moving Robots Based on Microcontroller Control Jie TANG and Xiao-min LIU

MWGen: A Mini World Generator

Combined impacts of configurational and compositional properties of street network on vehicular flow

Advanced PMA Capabilities for MCM

UAV-based monitoring of pedestrian groups

Journal of Chemical and Pharmaceutical Research, 2016, 8(6): Research Article. Walking Robot Stability Based on Inverted Pendulum Model

Sharp Shooting: Improving Basketball Shooting Form

Bicycle Theft Detection

Smart Cars for Safe Driving

Generation of See-Through Baseball Movie from Multi-Camera Views

Cricket umpire assistance and ball tracking system using a single smartphone camera

Automated Proactive Road Safety Analysis

Sensing and Modeling of Terrain Features using Crawling Robots

The Application of Pedestrian Microscopic Simulation Technology in Researching the Influenced Realm around Urban Rail Transit Station

Deakin Research Online

Qualification Document for RoboCup 2016

RECENTLY, various humanoid robots have been

CS 4649/7649 Robot Intelligence: Planning

Video Based Accurate Step Counting for Treadmills

Simulation and mathematical modeling for racket position and attitude of table tennis

Pedestrian crossing aid device for the visually impaired

CS 4649/7649 Robot Intelligence: Planning

Learning of Cooperative actions in multi-agent systems: a case study of pass play in Soccer Hitoshi Matsubara, Itsuki Noda and Kazuo Hiraki

A Bag-of-Gait Model for Gait Recognition

Using sensory feedback to improve locomotion performance of the salamander robot in different environments

High-Speed Batting Using a Multi-Jointed Manipulator

Fault Diagnosis based on Particle Filter - with applications to marine crafts

Motion Control of a Bipedal Walking Robot

Open Research Online The Open University s repository of research publications and other research outputs

Performance of Fully Automated 3D Cracking Survey with Pixel Accuracy based on Deep Learning

Football Pass Prediction using Player Locations

The Sweaty 2018 RoboCup Humanoid Adult Size Team Description

A System Development for Creating Indoor Floor Plan of Walking Route using Video and Movement Data 1 2

AutonoVi-Sim: Modular Autonomous Vehicle Simulation Platform Supporting Diverse Vehicle Models, Sensor Configuration, and Traffic Conditions

Evaluation of the Performance of CS Freiburg 1999 and CS Freiburg 2000

ZSTT Team Description Paper for Humanoid size League of Robocup 2017

Algorithm for Line Follower Robots to Follow Critical Paths with Minimum Number of Sensors

ICS 606 / EE 606. RoboCup and Agent Soccer. Intelligent Autonomous Agents ICS 606 / EE606 Fall 2011

Analysis and realization of synchronized swimming in URWPGSim2D

Predicting Human Behavior from Public Cameras with Convolutional Neural Networks

A Generalised Approach to Position Selection for Simulated Soccer Agents

SHUFFLE TURN OF HUMANOID ROBOT SIMULATION BASED ON EMG MEASUREMENT

IEEE RAS Micro/Nano Robotics & Automation (MNRA) Technical Committee Mobile Microrobotics Challenge 2016

Research on the Sealing Detection Technology for No Leak Detection Interface Specimen Yingjun Huanga, Xudong Liaob, Guoyun Baic, Tao Chend, Miao Loue

Recognition of Tennis Strokes using Key Postures

EXPERIMENTAL RESULTS OF GUIDED WAVE TRAVEL TIME TOMOGRAPHY

Titelbild. Höhe: 13cm Breite: 21 cm

Transformer fault diagnosis using Dissolved Gas Analysis technology and Bayesian networks

Problem Solving as Search - I

PEDESTRIAN behavior modeling and analysis is

Application of Dijkstra s Algorithm in the Evacuation System Utilizing Exit Signs

/435 Artificial Intelligence Fall 2015

Look Up! Positioning-based Pedestrian Risk Awareness. Shubham Jain

siot-shoe: A Smart IoT-shoe for Gait Assistance (Miami University)

Sensors & Transducers Published by IFSA Publishing, S. L., 2017

Computer Aided Drafting, Design and Manufacturing Volume 26, Number 2, June 2016, Page 53. The design of exoskeleton lower limbs rehabilitation robot

Investigation of Gait Representations in Lower Knee Gait Recognition

Heart Rate Prediction Based on Cycling Cadence Using Feedforward Neural Network

Visual Traffic Jam Analysis Based on Trajectory Data

1. Task Definition. Automated Human Gait Recognition Ifueko Igbinedion, Ysis Tarter CS 229, Fall 2013

Vibration Analysis and Test of Backup Roll in Temper Mill

THE PRESSURE SIGNAL CALIBRATION TECHNOLOGY OF THE COMPREHENSIVE TEST

The Incremental Evolution of Gaits for Hexapod Robots

High-Resolution Measurement-Based Phase-Resolved Prediction of Ocean Wavefields

Evaluation of Gait Recognition. Synonyms. Definition. A Panoramic View of Performance. Introduction

Robot motion by simultaneously wheel and leg propulsion

Robust Enhancement of Depth Images from Depth Sensors

A climbing motion recognition method using anatomical information for screen climbing games

Intelligent Decision Making Framework for Ship Collision Avoidance based on COLREGs

UNDERWATER DRONES CONTROL TOWER UNDERWATER DRONES CONTROL TOWER

YAN GU. Assistant Professor, University of Massachusetts Lowell. Frederick N. Andrews Fellowship, Graduate School, Purdue University ( )

Implementation of Height Measurement System Based on Pressure Sensor BMP085

67. Sectional normalization and recognization on the PV-Diagram of reciprocating compressor

IN-VEHICLE PEDESTRIAN DETECTION USING STEREO VISION TECHNOLOGY

Modelling Approaches and Results of the FHDO Biomedical Computer Science Group at ImageCLEF 2015 Medical Classification Task

Object Recognition. Selim Aksoy. Bilkent University

Dynamic Characteristics of the End-effector of a Drilling Robot for Aviation

Research on Key Technologies of Small Underwater Robots

PREDICTING the outcomes of sporting events

u = Open Access Reliability Analysis and Optimization of the Ship Ballast Water System Tang Ming 1, Zhu Fa-xin 2,* and Li Yu-le 2 " ) x # m," > 0

Generation of Robot Motion Based on Measurement of Human Movement. Susumu Sakano 1, Satoru Shoji 1

Compensator Design for Speed Control of DC Motor by Root Locus Approach using MATLAB

Transcription:

21 12 2017 12 Electri c Machines and Control Vol. 21 No. 12 Dec. 2017 SLAM 1 1 2 1 1. 150080 2. 100190 : 3 ( SLAM), SLAM V2,RTAB-Map DVO SLAM, 3 SLAM SLAM, TUM ICL-NUIM, SLAM, SLAM, SLAM :, SLAM V2;, DVO SLAM;, RTAB-Map : ; ; ; ; DOI: 10. 15938 /j. emc. 2017. 12. 008 TP 13 A 1007-449X 2017 12-0060- 06 Evaluation of the depth camera based SLAM algorithms MAN Chun-tao 1 CAO Miao 1 2 LI Wei 1 1. School of Automation Harbin University of Science and Technology Harbin 150080 China 2. Key Laboratory of Complex System and Intelligence Science Institute of Automation China Academy of Sciences Beijing 100190 China Abstract Three typical depth camera based simultaneous localization and mapping SLAM algorithms including SLAM V2 RTAB-Map and DVO SLAM whose theories and features were introduced. By using two open-source SLAM datasets including TUM dataset and ICL-NUIM dataset the above three SLAM algorithms were evaluated and the index included the accuracy performance and robustness of the SLAM algorithms. The results of the experiments demonstrate that SLAM V2 is chosen when accuracy and robustness are prior to speed DVO SLAM is chosen when speed and robustness are prior to accuracy RTAB-Map is chosen when speed and accuracy are prior to robustness. Keywords SLAM algorithm algorithm evaluation visual odometry mapping depth camera 0 2017-07 - 03 : ( F2016027) : ( 1965 ),,,, ; : ( 1992 ),,, ; ( 1982 ),,,,

12 SLAM 61 1 3 SLAM SLAM V2 1-2 RTAB-Map DVO SLAM 3 Kinect Xtion Pro Live 1. 1 SLAM V2 SLAM SLAMV2 SLAM Daniel 1 SLAM SLAM 3 SLAM SLAM SLAM RGB- SLAM D SLAM SLAM 4-7 SLAM SLAM SLAM SLAM SLAM Sturm 8 SLAM TUM 1 SLAM V2 Mur-Artal 9 Fig. 1 Schematic overview of SLAM V2 SLAM 3 Yousif 10 1. 1. 1 SLAM SLAM SLAM Kinect SLAM Endres 11 SLAM TUM SLAM OpenCV 12 SLAM SLAM SLAM 3 SLAM 13 SLAM V2 1. 1. 2 SLAM RTAB-Map DVO SLAM SLAM TUM ICL-NUIM SLAM SLAM g 2 o 14

62 21 SLAM 1. 1. 3 ground truth ground truth SLAM SLAM SLAM ICL-NUIM OctoMap 15 TUM 1 SLAM 1. 2 RTAB-Map 2 RTAB-Map SLAM ICL-NUIM SLAM TUM Michaud 16 3 SURF 2. 2 t t ground truth SLAM F = Q -1 SP Q ground truth P 1. 3 DVO SLAM DVO SLAM Horn 22 17 - SLAM 18 Christian 19 P Q S 1 2 t SLAM ground truth F t 3 g 2 o 2 SLAM V2 RTAB-Map DVO SLAM 3 SLAM TUM 20 ICL-NUIM SLAM V2 RT- 21 SLAM AB-Map 1. 7 1. 9 3 ICL-NUIM DVO SLAM 1 Intel Core i7 SLAM V2 RTAB-Map - 7500 2. 4GHz 16GB 2 5 6. 4 SLAM V2 1 2. 3 RTAB-Map 2 DVO SLAM 3 2. 1 360 ground truth F max 1 2 1 2 m 1 SLAM V2 RTAB-Map 2 TUM DVO SLAM SLAM SLAM V2 RTAB-Map TUM xyz SLAM

12 SLAM 63 RTAB-Map DVO SLAM DVO SLAM RGB- 1 SLAM TUM D SLAM Table 1 Accuracy of the SLAM algorithm with SLAM V2 respect to the TUM dataset SLAM 1 RGB- D SLAM SLAM SLAM t SLAM t SLAM t - t SLAM Table 2 Accuracy of the SLAM algorithm with 3 4 s respect to the ICL-NUIM dataset 1 TUM DVO SLAM SLAM V2 SLAM V2 RTAB-Map 0. 2 0. 5 2 ICL-NUIM DVO SLAM SLAM V2 RTAB-Map 1. 1 2. 9 2 3 fr1_360 0. 377 0. 223 0. 372 fr1_desk2 0. 191 0. 246 0. 31 fr1_floor 0. 211 0. 105 0. 65 fr1_plant 0. 148 0. 233 0. 202 fr1_room 0. 389 0. 146 0. 363 fr1_teddy 0. 161 0. 368 0. 598 0. 246 0. 22 0. 416 2 SLAM ICL-NUIM 1 2 3 room0 0. 216 0. 281 0. 906 room1 0. 354 0. 18 1. 233 room2 0. 187 0. 259 1. 118 room3 0. 264 0. 07 1. 84 0. 255 0. 198 1. 274 2 RTAB-Map Fig. 2 Robustness of the RTAB-Map algorithm SLAM 2. 4 3 4 s SLAM 1 TUM DVO SLAM SLAM V2 SLAM RTAB-Map 0. 2 0. 5 2 ICL-NUIM SLAM V2 RTAB-Map 1. 1 2. 9 DVO SLAM SLAM SLAM

64 21 SLAM SLAM 5 5 3 3 Table 3 SLAM ICL-NUIM Runtime of the SLAM algorithm with respect to the ICL-NUIM dataset 1 2 3 room0 264 101 297 room1 131 49 104 room2 75 29 82 room3 146 42 214 154 60 174 4 Table 4 SLAM TUM Runtime of the SLAM algorithm with respect to the TUM dataset 1 2 3 fr1_360 179 102 37 fr1_desk2 148 65 30 fr1_floor 313 102 78 fr1_plant 293 87 43 fr1_room 327 146 58 fr1_teddy 335 116 68 266 103 52 Map fr1_360 fr1_desk2 fr1_room fr1_teddy room0 room3 2 a TUM fr1_desk2 SLAM 2 b 2 c 2 d 1 SLAM TUM fr1 _360 fr1 _room fr1 _teddy 2 e 2 f SLAM V2 2 SLAM 3 RTAB-Map 2 RTAB- SLAM ICL-NUIM room0 room3 SLAM SLAM 2 RTAB- RTAB-Map Map 3 SLAM SLAM RTAB-Map DVO SLAM SLAM 1 RGB- D SLAM V2 DVO SLAM 100% 2 RTAB-Map 6 10 SLAM V2 DVO SLAM RTAB-Map Table 5 5 SLAM Robustness of the SLAM algorithm 1 2 3 TUM 6 /6 3 /6 6 /6 ICL-NUIM 4 /4 3 /4 4 /4 3 SLAM 1 SLAM V2 SIFT 22 2 DVO SLAM SLAM DVO SLAM SLAM RTAB-Map SLAM SLAM SLAM

12 SLAM 65 SLAM 11 ENDRES F SLAM 1 177. SLAM SLAM SLAM 1. 1987 726. J. 2017 21 6 104. MAN Chuntao CAO Miao Cao Yongcheng et al. Tracking and obstacle avoidance strategy for a bionic robot-rat J. Electric Machines and Control 2017 21 6 104. 2. J. probabilistic flexible and compact 3D map representation for robotic systems C / / Proc. of the ICRA Workshop on Best Prac- 2007 11 1 79. GUO Xu XIONGRong HU Xiehe. Motion predictive control of omni-directional mobile robot J. Electric Machines and Control Manipulation. tice in 3D Perception and Modeling for Mobile 2010. 2007 11 1 79. 3 CADENA C CARLONE L CARRILLO H et al. Past present and future of simultaneous localization and mapping toward the robustperception age J. IEEE Transactions on Robotics 2016 32 6 1309. 4 THRUN S. Robotic mapping a survey M / / Exploring artificial intelligence in the new millennium. Morgan Kaufmann Publishers Inc. 2002 2002. 5 NUCHTER A LINGEMANN K HERTZBERG J et al. 6 D SLAM with approximate data association C / / International Conference on Advanced Robotics 2005. IEEE 2005 242. 6 GRISETTI G GRZONKA S STACHNISS C et al. Efficient estimation of accurate maximum likelihood maps in 3D C / /Ieee /rsj International Conference on Intelligent Robots and Systems. IEEE 2007 3472. 7 FRESE U LARSSON P DUCKETT T. A multilevel relaxation algorithm for simultaneous localization and mapping J. IEEE Transactions on Robotics 2005 21 2 196. 8 STURM J ENGELHARD N ENDRES F et al. A benchmark for theevaluation of SLAM system C / / Ieee /rsj International Conference on Intelligent Robots and Systems. IEEE 2012 573. 9 MURARTAL R TARDOS J D. ORB-SLAM2 an open-source SLAM system for monocular stereo and cameras J. IEEE Transactions or Robotics 2016 99 1. 10 YOUSIF K BABHADIASHAR A HOSEINNEZHAD R. Realtime registration and mapping in texture-less environments using ranked order statistics C / / Ieee /rsj International Conference on Intelligent Robots and Systems. IEEE 2014 2654. HESS J STURM J et al. 3-D mapping with an RGB- D camera J. IEEE Transactions on Robotics 2017 30 12 BRADSKI G KAEHLER A. Learning opencv computer vision in C + + with the opencv library M. O'Reilly Media Inc. 2013. 13 FISCHLER M A BOLLES R C. Random sample consensus a paradigm for model fitting with applications to image analysis and automated cartography J. Readings in Computer Vision 14 K MMERLE R GRISETTI G STRASDAT H et al. G2o a general framework for graph optimization C / /IEEE International Conference on Robotics and Automation. IEEE 2011 e43478. 15 WURM K M HORNUNG A BENNEWITZ M et al. OctoMap a 16 LABB M MICHAUD F. Online global loop closure detection for large-scale multi-session graph-based SLAM C / / International Conference on Intelligent Robots and Systems. IEEE 2014 2661. 17 KERL C STURM J CREMERS D. Robust odometry estimation for cameras C / / IEEE International Conference on Robotics and Automation. IEEE 2013 3748. 18 HENRY P KRAININ M HERBST E et al. mapping Using kinect-style depth cameras for dense 3D modeling of indoor environments M / / Experimental Robotics. Springer Berlin Heidelberg 2014 647. 19 STUCKLER J BEHNKE S. Integrating depth and color cues for dense multi-resolution scene mapping using cameras C / / Multi-sensor Fusion and Integration for Intelligent Systems. IEEE 2013 162. 20 ENDRES F HESS J ENGELHARD N et al. Anevaluation of the SLAM system C / /IEEE International Conference on Robotics and Automation. IEEE 2012 1691. 21 HANDA A WHELAN T MCDONALD J et al. A benchmark for visual odometry 3D reconstruction and SLSM C / / IEEE International Conference on Robotics and Automation. IEEE 2014 1524. 22. J. 2014 18 11 78. WANG Fochi YAN Kang ZHANG Chongyuan et al. Identifying insultor hydrophobicity by image analysis and nueral network J. Electric Machines and Control 2014 18 11 78. ( : )