TaeHyoung Kim( ) Review 2 Intention-Aware Online POMDP Planning - - PowerPoint PPT Presentation

▶
taehyoung kim review
SMART_READER_LITE
LIVE PREVIEW

TaeHyoung Kim( ) Review 2 Intention-Aware Online POMDP Planning - - PowerPoint PPT Presentation

Dec 16, 2022 162 likes •724 views

Intention-Aware Online POMDP Planning for Autonomous Driving in a Crowd Bai, Haoye, et al. ICRA 2015 TaeHyoung Kim( ) Review 2 Intention-Aware Online POMDP Planning for Autonomous Driving in a Crowd Bai, Haoye, et al. ICRA 2015

slide-1
SLIDE 1

Intention-Aware Online POMDP Planning for Autonomous Driving in a Crowd

Bai, Haoye, et al. ICRA 2015

TaeHyoung Kim(김태형)

slide-2
SLIDE 2

2

Review

slide-3
SLIDE 3

3

Intention-Aware Online POMDP Planning for Autonomous Driving in a Crowd

Bai, Haoye, et al. ICRA 2015

TaeHyoung Kim(김태형)

slide-4
SLIDE 4

4

Abstract

  • Goal: Autonomous driving among many

pedestrians effectively and safely.

  • Main contribution:
  • Online planning
  • Consider long-term effect of action

C.f.) Reactive control

slide-5
SLIDE 5

5

Reactive controller

slide-6
SLIDE 6

6

Reactive Control

  • Two state for pedestrian behavior
  • Stays on side walk
  • Crosses the road

Belief ( p , 1-p )

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-7
SLIDE 7

7

Reactive Control

  • For time n, Belief~ (0.51,0.49)

Accelerate

slide-8
SLIDE 8

8

Reactive Control

  • For time n, Belief~ (0.51,0.49)

Accelerate

slide-9
SLIDE 9

9

Reactive Control

  • For time n+ 1, Belief~ (0.35,0.65)

Decelerate

slide-10
SLIDE 10

10

Reactive Control

  • For time n+ 1, Belief~ (0.35,0.65)

Too late..

slide-11
SLIDE 11

11

System overview

slide-12
SLIDE 12

12

System models

  • Vehicle Model
  • Position ,
  • Orientation
  • I nstantaneous speed
  • Pedestrian Model
  • Position ,
  • I nstantaneous speed
  • Goal (intention - Explained later)
  • Sensor Model
  • Vehicle position, speed
  • Positions of all pedestrians
slide-13
SLIDE 13

13

System Overview

  • For every time step,
  • Belief tacking
  • Path planning
  • Speed planning

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-14
SLIDE 14

14

Belief Tracker

slide-15
SLIDE 15

15

Sub-goal Concept

  • From human science studies.
  • Sub-goal
  • points in a space that pedestrians are walking

toward

  • landmarks of environment

Ikeda, Tetsushi, et al. "Modeling and prediction of pedestrian behavior based on the sub-goal concept." Robotics (2013): 137.

slide-16
SLIDE 16

16

Belief of Pedestrians’ intention

  • Belief of Pedestrians’ intention
  • Probability distribution for each sub-goals

Belief

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-17
SLIDE 17

17

Pedestrian model

  • Pedestrian Model
  • Position ,
  • I nstantaneous velocity,
  • Goal

The Highest possible sub-goal position in Belief

slide-18
SLIDE 18

18

Belief Tracker

  • Using observed pedestrian’s movement
  • Bayer’s rule

Current position Previous position Velocity, goal New belief

slide-19
SLIDE 19

19

Belief Tracker

  • Use Belief
  • Utilized in path planning & speed planning
  • Up to 7 Pedestrians

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-20
SLIDE 20

20

Path Planning

slide-21
SLIDE 21

21

Path planning

  • Grid World + Grid search
  • Path, : , , , …
  • Path cost,

Static obstacle Pedestrians Smoothness

Potential Field :

slide-22
SLIDE 22

22

Path Planning – Grid Search

  • Grid Search
  • Regular A*
  • Does not consider non-holonomic constraint

Petereit, Janko, et al. "Application of Hybrid A* to an autonomous mobile robot for path planning in unstructured outdoor environments." Robotics; Proceedings of ROBOTIK 2012; 7th German Conference on. VDE, 2012.

slide-23
SLIDE 23

23

Path Planning – Hybrid A*

  • Hybrid A*
  • For each cell, also contains continuous position.

Petereit, Janko, et al. "Application of Hybrid A* to an autonomous mobile robot for path planning in unstructured outdoor environments." Robotics; Proceedings of ROBOTIK 2012; 7th German Conference on. VDE, 2012.

slide-24
SLIDE 24

24

Path Planning – Hybrid A* detail

  • I n detail procedure

Initial situation Open set Close set

slide-25
SLIDE 25

25

Path Planning – Hybrid A* detail

  • I n detail procedure

Select node from open set to expand Open set Close set

slide-26
SLIDE 26

26

Path Planning – Hybrid A* detail

  • I n detail procedure

Expand node Open set Close set

slide-27
SLIDE 27

27

Path Planning – Hybrid A* detail

  • I n detail procedure

Select one point in each cell Open set Close set

slide-28
SLIDE 28

28

Path Planning – Hybrid A* detail

  • I n detail procedure

Open set Close set Select node from open set to expand

slide-29
SLIDE 29

29

Path Planning – Hybrid A* detail

  • I n detail procedure

Open set Close set Expand node

slide-30
SLIDE 30

30

Path Planning – Hybrid A* detail

  • I n detail procedure

Open set Close set Select one point in each cell

slide-31
SLIDE 31

31

Path Planning – Hybrid A* detail

  • I n detail procedure

Open set Close set Select node from open set to expand

slide-32
SLIDE 32

32

Path Planning – Hybrid A* detail

  • I n detail procedure

Open set Close set Expand & Select one point in each cell

slide-33
SLIDE 33

33

Path Planning – Hybrid A* detail

  • I n detail procedure

Open set Close set Find continuous path

slide-34
SLIDE 34

34

Path Planning

  • Set current steering angle
  • Situation is continuously changing

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-35
SLIDE 35

35

Speed Planner

  • Collision Avoidance
slide-36
SLIDE 36

36

Speed planning

  • Assumption
  • Pedestrian walks toward it’s goal
  • Pedestrian speed is constant during planning

cycle

  • Perfect sensor
slide-37
SLIDE 37

37

Collision Avoidance

  • Select Acceleration
  • Action: ACCEL. / MAI NTAI N / DECEL.
  • Utilize
  • Path from path planner
  • Belief from belief tracker – For penalty

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA) 2015 IEEE International Conference on IEEE 2015

slide-38
SLIDE 38

38

Framework – Online POMDP

  • POMDP model
  • Vehicle(, , , )
  • Pedestrians , , , up to 7
  • Sensor model: discretized values
  • Action: Acceleration
  • (ACCELERATE, MAI NTAI N, DECELERATE)
  • Rewards & Penalties: Next Page…

Current situation

slide-39
SLIDE 39

39

Framework – Online POMDP

  • Reward
  • Large reward around Goal

 to reach the destination

  • Penalties
  • Large penalty for approaching the pedestrians

 for safe

  • Slow speed

 For driving at a higher speed

  • Accelerate and Decelerate actions

 For smooth driving

slide-40
SLIDE 40

40

Framework – Online POMDP

  • Online POMDP
  • Only finite horizon
  • Scenario sampling
slide-41
SLIDE 41

41

Framework – Online POMDP

  • Online POMDP procedure

Current belief : vehicle state, pedestrian beliefs

slide-42
SLIDE 42

42

Framework – Online POMDP

  • Online POMDP procedure

Accelerate Decelerate Maintain

slide-43
SLIDE 43

43

Framework – Online POMDP

  • Online POMDP procedure

Accelerate Decelerate Maintain Reward Reward Reward

slide-44
SLIDE 44

44

Framework – Online POMDP

  • Online POMDP procedure

Accelerate Decelerate Maintain

z1 z2 z3 z1 z2 z3 z1 z2 z3

slide-45
SLIDE 45

45

Framework – Online POMDP

  • Online POMDP procedure

Accelerate Decelerate Maintain

z1 z2 z3 z1 z2 z3 z1 z2 z3

Scenario

slide-46
SLIDE 46

46

Framework – Online POMDP

  • The problem is scenarios grow exponentially

Accelerate Decelerate Maintain

z1 z2 z3 z1 z2 z3 z1 z2 z3

slide-47
SLIDE 47

47

Framework – Online POMDP

  • The problem is scenarios grow exponentially
slide-48
SLIDE 48

48

Framework – Online POMDP

  • Online POMDP procedure
  • Random sampling of observations

Accelerate Decelerate Maintain

z1 z2 z3 z1 z2 z3 z1 z2 z3

Finite horizon

slide-49
SLIDE 49

49

Framework – Online POMDP

  • Online POMDP procedure

Finite horizon sampling The best action

slide-50
SLIDE 50

50

Framework – Online POMDP

  • Utilize finite horizon scenarios
  • Consider long-term effect of the current action
  • Execute current action

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-51
SLIDE 51

51

Demo video

slide-52
SLIDE 52

52

Result

  • Demo video

Bai, Haoyu, et al. "Intention-aware online POMDP planning for autonomous driving in a crowd." Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.

slide-53
SLIDE 53

53

Pros and cons

slide-54
SLIDE 54

54

Pros and cons

  • Pros
  • Seems somewhat success.
  • Tries to anticipate future.
  • There is room for development. (Deep learning)
  • Cons
  • Sub-goal concept is somewhat restricted.
  • The pedestrians should behave normally.
  • Decision quality trade off with computation

time.

slide-55
SLIDE 55

55

  • Q&A