Sensing Ability Based Exploration Date : 2019.12.03 Presenter : - - PowerPoint PPT Presentation

▶
sensing ability based exploration
SMART_READER_LITE
LIVE PREVIEW

Sensing Ability Based Exploration Date : 2019.12.03 Presenter : - - PowerPoint PPT Presentation

Mar 23, 2023 410 likes •771 views

Sensing Ability Based Exploration Date : 2019.12.03 Presenter : Duckyu Choi ID : 20194554 0. Background Representing Space As for most sensors the perception stops at surfaces, hollow spaces or narrow pockets can sometimes not be explored

slide-1
SLIDE 1

Date : 2019.12.03 Presenter : Duckyu Choi ID : 20194554

Sensing Ability Based Exploration

slide-2
SLIDE 2
  • 0. Background

Representing Space Purpose of Exploration

1

As for most sensors the perception stops at surfaces, hollow spaces or narrow pockets can sometimes not be explored with a given setup. This residual space denoted by

slide-3
SLIDE 3
  • 0. Background

Exploration

  • Map is used for both, collision free navigation and determination of the

exploration progress

2 Mapping

Motion Control

Localization

Exploration

slide-4
SLIDE 4
  • 0. Background

Previous Exploration Method

  • Frontier Based Exploration is one of them
  • To gain the most new information about the world, move to the boundary

between known space and unknown space

3

Goal Unknown area Robot C

Observed area

Frontier

slide-5
SLIDE 5
  • 0. Background

Previous Exploration Method

  • Frontier Based Exploration is one of them
  • To gain the most new information about the world, move to the boundary

between known space and unknown space

4

Goal Frontier Map update

slide-6
SLIDE 6
  • 0. Background

Octomap, Occupancy map

  • “OctoMap: an efficient probabilistic 3D mapping framework based on
  • ctrees, Auton Robot (2013) 34:189–206”

5

slide-7
SLIDE 7
  • 1. Receding Horizon "Next–Best–View"

Planner for 3D Exploration

slide-8
SLIDE 8
  • 1. NBVP

Basic Framework Paths are only planned through known free space 𝑊free, thus providing collision–free navigation

7

RRT Tree Generation Tree Branch Selection Receding Horizon

Model Predictive Control

slide-9
SLIDE 9
  • 1. NBVP

Receding Horizon Planning (RHP)

  • Only the first waypoint is executed when the robot moves.
  • The map is updated as more grids are explored, and the path is re-

planned if necessary.

  • Able to plan a smooth path where waypoints can be located on any

position on the edge of grids without linear interpolation, which may not work for a cost function that includes nonlinear factors.

8

slide-10
SLIDE 10
  • 1. NBVP

Summation of Gain – Selecting Best Qualified Tree

9

3 2 1 2 1 3

+ new gain + new gain

slide-11
SLIDE 11
  • 1. NBVP

Quality - collected information gain

10

slide-12
SLIDE 12
  • 1. NBVP

11

slide-13
SLIDE 13
  • 1. NBVP

12

slide-14
SLIDE 14
  • 1. NBVP

Computational Complexity

13

RRT complexity Occupancy map complexity with 1/𝒔𝟒 scale Gain computation complexity

V: volume to explore r: resolution of occupancy map 𝒆𝒏𝒃𝒚

𝒒𝒎𝒃𝒐𝒐𝒇𝒔: sensor range

𝑶𝑼: number of nodes in the tree

slide-15
SLIDE 15

Bridge Simulation Experiment

  • 1. NBVP

Indoor Simulation Experiment

14

slide-16
SLIDE 16
  • 1. NBVP

Real World Experiment

15

slide-17
SLIDE 17
  • 2. Uncertainty–aware Receding Horizon

Exploration and Mapping using Aerial Robots

slide-18
SLIDE 18
  • 2. UEP

Key requirements for the VIO to perform robustly

  • must reobserve landmarks with good confidence
  • better to follow trajectories that appropriately excite the inertial

sensors

This has two effects

  • improving the location estimate of the features
  • improving the pose estimate of the robot due to the statistical

correlations that link the vehicle to the features.

17

slide-19
SLIDE 19
  • 2. UEP

Especially when the robot explores an unknown environment, new features are initialized into the map. This imposes the need to reobserve previous features in order to reduce the growth in localization and mapping error

18

slide-20
SLIDE 20
  • 2. UEP

Two–Step Planner

  • First, receding horizon planner
  • Second, belief space–based planner

19

slide-21
SLIDE 21
  • 2. UEP

Two–Step Planner

  • S’’ has selected by receding horizon planner
  • S’’ be the goal for nested second planner

20

slide-22
SLIDE 22
  • 2. UEP

Two–Step Planner

  • S’’ has selected by receding horizon planner
  • S’’ be the goal for nested second planner

21

slide-23
SLIDE 23
  • 2. UEP

Two–Step Planner

  • S’’ has selected by receding horizon planner
  • S’’ be the goal for nested second planner

22

slide-24
SLIDE 24
  • 2. UEP

Two–Step Planner

  • Make a local volume and sample the random tree
  • Evaluating by D-optimality, generate nested path

23

Random Sampling In local volume

slide-25
SLIDE 25
  • 2. UEP

Two–Step Planner

  • Make a local volume and sample the random tree
  • Evaluating by D-optimality, generate nested path

24

Nested Path Generation

slide-26
SLIDE 26
  • 2. UEP

25

slide-27
SLIDE 27
  • 2. UEP

26

Gain Function(1st planner)

P(m) := probability of occupied voxel

slide-28
SLIDE 28
  • 2. UEP

27

Gain Function(2nd planner)

Σp,f : Derived pose and tracked landmarks covariance matrix of robot and feature state, in the paper they used EKF covariance which is used with ROVIO image patch lp, lf : Dimensions of pose, robot state, features state

slide-29
SLIDE 29
  • 2. UEP

Computational Complexity

28

slide-30
SLIDE 30
  • 2. UEP

29

slide-31
SLIDE 31
  • 2. UEP

30

slide-32
SLIDE 32
  • 2. UEP

31

slide-33
SLIDE 33
  • 3. Conclusion

Index NBVP UEP

Similar Point

Space Representation Occupancy map(Octomap) Path Randomness RRT based sampling, Frequent back & forth movement

Different Point

Considering Localization No YES(reobservation gain) Path Complexity Low High(second path planner) Exploration Time Better than FE Worse time performance

32

slide-34
SLIDE 34
  • 4. Quiz

1) NBV-planner have visibility element on gain function. (T/F) 2) UEP is short name of ‘Unsaturation Aware Planning’. (T/F)

33

slide-35
SLIDE 35

Thank you