Online Path Planning for Autonomous Underwater Vehicles in Unknown - PowerPoint PPT Presentation

Online Path Planning for Autonomous Underwater Vehicles in Unknown Environments Juan David Hernandez, Eduard Vidal, Guillem Vallicrosa, Enric Galceran, and Marc Carreras ICRA 2015 Presented by Youngki Kwon

Contents ● Introduction ● Main contribution ● Background ● Anytime path planning ● Lazy collision evaluation ● Octree-based map ● Main approach ● Framework ● Motion planning ● Experiment and result ● Summary 3

INTRODUCTION 4

Introduction ● AUVs must operate in unknown, cluttered and dynamic environments ● AUVs are more exposed to collisions ● Drift effects on the position estimated by navigation system ● Online capabilities of path planner make AUVs overcome global position inaccuracy 5

MAIN CONTRIBUTION 6

Main Contribution ● Combination of ideas from lazy collision evaluation and anytime path planning algorithm ● For online path planning ● incrementally built octree-based representation of the environment ● The experimental evaluation of the resulting planning framework using the SPARUS-II AUV in a real-world setting 7

BACKGROUND 8

Anytime Path Planning ● Overall approach ● Quickly find a solution that is feasible, but not necessarily optimal ● Exploit execution time to incrementally improve towards optimal solution ● Desired properties ● Form of completeness guarantees ● Asymptotic optimality given more computation time Anytime RRT* - Matthew Walter’s slide 9

Lazy Collision Evaluation ● Only check for collisions when we have to Check the collision after building the roadmap 11

Octree-based Map Techniques for 3d mapping - Wolfram burgard’s slide 12

Octree-based Map ● Octomap Techniques for 3d mapping - Wolfram burgard’s slide 13

MAIN APPROACH 14

Framework ● To solve start-to-goal queries for an AUV in an unknown environment ● Framework for path planning online for AUVs 15

Anytime Approach - Under the time constraint, current T is left image - Assume T is valid 𝒓 𝒉𝒑𝒃𝒎 Time limit 𝑺𝒑𝒑𝒖 16

Anytime Approach - Traverse T, check the collision, cut the subtree in collision 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 Lazy Collision Evaluation 17

Anytime Approach - Sample configuration 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 18

Anytime Approach - Extension in RRT* 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 19

Anytime Approach - Extension in RRT* 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 20

Anytime Approach - Extension in RRT* 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 21

Anytime Approach - Extension in RRT* 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 22

Anytime Approach - Extension in RRT* 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 23

Anytime Approach - Find solution 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 24

Anytime Approach - After Satisfying stop condition 𝒓 𝒉𝒑𝒃𝒎 𝑺𝒑𝒑𝒖 𝑺𝒑𝒑𝒖 25

EXPERIMENT & RESULT 26

Experiment & Result ● AUV use ● Pressure sensor, Doppler velocity log, IMU measurement, GPS ● Five Echosounders ● Experiments in both simulation and real world 27

Experiment & Result - Experiment in simulation 28

Experiment & Result - Experiment in real-world 29

SUMMARY 30

Summary ● Online capabilities of path planner make AUVs overcome global position inaccuracy ● Combination of ideas from lazy collision evaluation and anytime path planning algorithm 31

Q & A 32