lidar based obstacle detection and collision .lidar based obstacle detection and collision avoidance

Download LiDAR based obstacle detection and collision .LiDAR based obstacle detection and collision avoidance

Post on 20-Aug-2018

212 views

Category:

Documents

0 download

Embed Size (px)

TRANSCRIPT

  • Bachelors Thesis

    CzechTechnicalUniversityin Prague

    F3 Faculty of Electrical EngineeringDepartment of Control Engineering

    LiDAR based obstacle detectionand collision avoidance in anoutdoor environment

    Jan PedotaCybernetics and robotics

    May 2016Supervisor: Ing. Milan Rollo, Ph.D.

  • Czech Technical University in Prague Faculty of Electrical Engineering

    Department of Control Engineering

    BACHELOR PROJECT ASSIGNMENT

    Student: Jan Pedota

    Study programme: Cybernetics and Robotics Specialisation: Systems and Control

    Title of Bachelor Project: LiDAR based obstacle detection and collision avoidance in outdoor environment

    Guidelines:

    1. Study the problematics of navigation based on laser rangefinder in unknown outdoor environment 2. Integrate essential sensors onto an autonomous unmanned ground vehicle (UGV) 3. Implement methods for sensory data processing and representation and generate obstacles for autonomous mobile vehicles 4. Adjust the planning algorithms of existing system for command&control of autonomous vehicles so that the vehicles can react to new obstacles and change their trajectories in real time 5. Verify the functionality of the resulting system in real environment

    Bibliography/Sources:

    [1] Weitkamp, C. ed. Lidar: range-resolved optical remote sensing of the atmosphere (Vol. 102). Springer Science & Business, 2006, (online: http://home.ustc.edu.cn/~522hyl/) [2] Hornung, A., Wurm, K.M., Bennewitz, M., Stachniss, C. and Burgard, W. OctoMap: An efficient probabilistic 3D mapping framework based on octrees. Autonomous Robots, 34(3), pp.189-206. 2013 [3] Vanneste, S., Bellekens, B. and Weyn, M., 3DVFH+: Real-Time Three-Dimensional Obstacle Avoidance Using an Octomap.

    Bachelor Project Supervisor: Ing. Milan Rollo, Ph.D.

    Valid until the summer semester 2016/17

    L.S.

    prof. Ing. Michael ebek, DrSc. Head of Department

    prof. Ing. Pavel Ripka, CSc. Dean

    Prague, February 11, 2016

  • Acknowledgement / DeclarationI would like to thank to my super-

    visor, Ing. Milan Rollo, Ph.D., forhis guidance, professional attitude andvaluable consultations and to Ing. Mar-tin Seleck for his constructive criticismand problem discussions. I would alsolike to thank to Tom Trafina, withwhom I conducted hardware exper-iments, and to my family for theirsupport during my studies.

    I declare that the presented workwas developed independently and thatI have listed all sources of informationused within it in accordance with themethodical instructions for observingthe ethical principles in the preparationof university theses.

    Prague, date 26.5.2016

    . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    iii

  • Abstrakt / AbstractTato prce se zabv problematikou

    scanovn neznmho okolnho pro-sted pomoc LiDARu, vytvenmmraen bod a OctoMap, detekc no-vch statickch pekek a pravouplnovacch algoritm pro vyhbn setmto pekkm. Pi tom tak popi-suje konstrukci pozemnho bezpilotnhovozidla a integraci potebnch senzorna toto vozidlo.

    Klov slova: LiDAR; IMU; RTKGPS; UGV; mrana bod; OctoMap;Point Cloud Library; RRT*; TacticalAgentFly; detekce pekek; plnovntrajektori.

    Peklad titulu: Detekce pekek po-moc LiDARu a pedchzen koliznm si-tuacm

    This thesis discusses the problematicsof unknown environment scanning us-ing LiDAR, point cloud and OctoMapconstruction, new static obstacles detec-tion and modification of current plan-ning algorithms for collision avoidance.It also describes construction of an au-tonomous ground vehicle and integra-tion of necessary sensors.

    Keywords: LiDAR; IMU; RTK GPS;UGV; pointcloud; OctoMap; PointCloud Library; RRT*; Tactical Agent-Fly; collision detection; path planning.

    iv

  • Contents /1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .11.1 Thesis outline . . . . . . . . . . . . . . . . . . . .1

    2 Typical methods of navigation . . . .22.1 Indoor navigation . . . . . . . . . . . . . . . .22.2 Outdoor navigation . . . . . . . . . . . . . .32.3 Pointcloud . . . . . . . . . . . . . . . . . . . . . . . .3

    3 Sensors and other devices . . . . . . . . .53.1 LiDAR. . . . . . . . . . . . . . . . . . . . . . . . . . . .53.2 IMU . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .53.3 RTK GPS . . . . . . . . . . . . . . . . . . . . . . . .63.4 3DR Radio . . . . . . . . . . . . . . . . . . . . . . .73.5 Microhard nVIP2400. . . . . . . . . . . . .73.6 ArduPilotMega. . . . . . . . . . . . . . . . . . .83.7 Toradex Iris and Colibri T30 . . . .9

    4 Rover platform . . . . . . . . . . . . . . . . . . . 104.1 Communication layer . . . . . . . . . . 104.2 Power . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

    5 Data processing and represen-tation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

    5.1 Initialization phase. . . . . . . . . . . . . 155.2 Time synchronization . . . . . . . . . . 15

    5.2.1 Time systems shift cor-rection . . . . . . . . . . . . . . . . . . . . 15

    5.2.2 Static time delay attransmission lines . . . . . . . . 16

    5.3 Data aggregation. . . . . . . . . . . . . . . 175.4 Calculation . . . . . . . . . . . . . . . . . . . . . 17

    5.4.1 Time interpolation . . . . . . . 175.4.2 Transformation. . . . . . . . . . . 17

    5.5 OctoMap . . . . . . . . . . . . . . . . . . . . . . . 175.5.1 OcTree . . . . . . . . . . . . . . . . . . . 17

    6 Obstacle detection . . . . . . . . . . . . . . . 206.1 Data preparation. . . . . . . . . . . . . . . 20

    6.1.1 Noise filtering . . . . . . . . . . . . 206.1.2 Downsampling . . . . . . . . . . . 216.1.3 Ground segmentation . . . . 21

    6.2 OctoMap updates and raytracing . . . . . . . . . . . . . . . . . . . . . . . . . . 23

    6.3 Collision detection . . . . . . . . . . . . . 247 Planning algorithms adjust-

    ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267.1 Tactical AgentFly . . . . . . . . . . . . . . 267.2 TAF modification . . . . . . . . . . . . . . 26

    7.2.1 Simple trajectory plan-ner . . . . . . . . . . . . . . . . . . . . . . . . 27

    7.2.2 RRT* trajectory plan-ner . . . . . . . . . . . . . . . . . . . . . . . . 29

    8 Experiments . . . . . . . . . . . . . . . . . . . . . . 328.1 Point cloud creation tests . . . . . 32

    8.1.1 RTK float solution im-precision . . . . . . . . . . . . . . . . . . 32

    8.1.2 RTK antenna offset . . . . . . 338.2 Improved point clouds . . . . . . . . . 35

    8.2.1 Difference betweenIMU and RTK coordi-nate system . . . . . . . . . . . . . . 35

    8.3 Collision avoidance. . . . . . . . . . . . . 368.3.1 OctoMap prepared in

    advance . . . . . . . . . . . . . . . . . . . 378.3.2 On the fly processing

    with collision avoidance . 409 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 449.1 Future work . . . . . . . . . . . . . . . . . . . . 44References . . . . . . . . . . . . . . . . . . . . . . . . 46

    A The attached CD-ROM . . . . . . . . . 49B Abbreviations . . . . . . . . . . . . . . . . . . . . . 50

    v

  • Tables / Figures4.1. Devices power consumption . . . 13 2.1. Point cloud - our office . . . . . . . . . . .4

    3.1. Velodyne Puck VLP-16 . . . . . . . . . .53.2. Microstrain 3DM-GX4-45. . . . . . . .63.3. Piksi RTK GPS . . . . . . . . . . . . . . . . . .73.4. 3DR Radios . . . . . . . . . . . . . . . . . . . . . .83.5. Microhard nVIP2400 . . . . . . . . . . . . .83.6. APM 2.6 . . . . . . . . . . . . . . . . . . . . . . . . . .93.7. Iris & Colibri T30 . . . . . . . . . . . . . . . .94.1. RC car and its chassis . . . . . . . . . 104.2. Connection scheme . . . . . . . . . . . . . 114.3. First floor bottom . . . . . . . . . . . . . . 114.4. First floor top . . . . . . . . . . . . . . . . . . 124.5. Rover second floor . . . . . . . . . . . . . 124.6. 5V and 12V distribution

    splitters . . . . . . . . . . . . . . . . . . . . . . . . . 134.7. LED indicator . . . . . . . . . . . . . . . . . . 144.8. Complete rover . . . . . . . . . . . . . . . . . 145.1. Point cloud creation process . . . 165.2. OcTree explanation . . . . . . . . . . . . 185.3. OctoMap different voxel sizes

    example . . . . . . . . . . . . . . . . . . . . . . . . . 186.1. PCL point cloud filtration . . . . . 216.2. PCL voxel grid down sam-

    pling enabled . . . . . . . . . . . . . . . . . . . 226.3. PCL voxel grid down sam-

    pling disabled. . . . . . . . . . . . . . . . . . . 226.4. Ground segmentation . . . . . . . . . . 236.5. OctoMap ray tracing . . . . . . . . . . . 246.6. OcTree example Letn . . . . . . . . . 257.1. TAF GUI . . . . . . . . . . . . . . . . . . . . . . . 267.2. Round NFZ avoidance . . . . . . . . . 287.3. Block NFZ avoidance . . . . . . . . . . 287.4. RRT algorithm explanation . . . 297.5. RRT* path planning - initial

    state . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307.6. RRT* path planning - path

    found . . . . . . . . . . . . . . . . . . . . . . . . . . . 318.1. Point cloud Charles square . . . . 328.2. Point cloud Strahov . . . . . . . . . . . . 338.3. RTK float problem - static . . . . 338.4. RTK float problem - dynamic . 348.5. Antenna offset illustration . . . . . 348.6. Scan around obstacle . . . . . . . . . . 358.7. Scan around obstacle 2 . . . . . . . . 358.8. Flat area photo. . . . . . . . . . . . . . . . . 368.9. Rectangle area photo . . . . . . . . . . 36

    vi

  • 8.10. RTK and IMU coordinatesystem difference 1 . . . . . . . . . . . . . 37

    8.11. RTK and IMU coordinatesystem difference 2 . . . . . . . . . . . . . 37

    8.12. Flat surface PC - first CA test . 388.13. Rectangle shaped PC - sec-

    ond CA test . . . . . . . . . . . . . . . . . . . . 388.14. Point cloud processing . . . . . . . . . 398.15. Flat area OctoM

Recommended

View more >