Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...•...

21
[CS686] Paper presentation 1 KAIST School of Computing Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order Bezier Curve and Local Optimal Reshaping Liang Yang et al. IROS 2015 Heechan Shin 2017.04.27 CS686

Transcript of Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...•...

Page 1: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Generation of Dynamically Feasible and Collision Free Trajectoryby Applying Six-order Bezier Curve and Local Optimal Reshaping

Liang Yang et al.

IROS 2015

Heechan Shin

2017.04.27

CS686

Page 2: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Contents

• Problem of the paper• Backgrounds• Framework• Methods• Result• Summary• Limitation• Quiz

2

Page 3: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Problems of the paper

• UAVs(or other robots) can follow a path, which is designed by previous path planner, with relatively slow speed.

• Piecewise linear paths are generated by RRT, RRT*, etc.

3

Page 4: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Backgrounds

• Curvature

4

Page 5: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Backgrounds

• Curvature

• Why is the curvature important?• Non-holonomic robots can’t go through piecewise linear path with high speed

5

Low speed High speed

Page 6: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Backgrounds

6

• G2-Continuity

• End point• Tangent• Curvature radius

Page 7: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Backgrounds

• Bezier curve• Using Bernstein polynomial• N-1 order Bezier curve with N points• Continuous curvature at joint of two curves

7

Page 8: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Framework

8

Page 9: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Framework

• Path planner : Guiding Attraction based Random Tree(GART)• Attraction to the goal & Repulsion by the obstacles

9

Page 10: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Methods

• Bezier curve based path smoothing

• Dangerous region finding

• Local optimal reshaping

10

Page 11: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Methods – Bezier curve based path smoothing

11

Six-order → 7 points

From path planner

From interpolation

Page 12: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Methods – Bezier curve based path smoothing

• How can we get interpolation points?

12

Page 13: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Methods – Bezier curve based path smoothing

13

• Tuning Rotation(TR)• Turning adjacent node until smooth

enough• Making the curve be consistent

Page 14: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Methods – Dangerous region finding

• “UAV simulations with dynamics show that the dangerous region”

• Performing fast simulation

• Obtaining dangerous regions (collision-expected)

14

Page 15: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Methods – Local optimal reshaping

• Nodes near dangerous region should be reshaped

15

Path length Curvature(1)

(1) Minimum distance from center of object𝒅𝒅𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 = 𝟒𝟒𝟒𝟒

(2),(3) Redistribution region

(2)

(3)

(4) Bezier curve

(4)

Page 16: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

• Curvature within maximum curvature after smoothing and reshaping• Not enough to their target curvature 0.1

Result

16

Page 17: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Result

17

• After Tuning Rotation, curvature becomes lower than 0.1

Page 18: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Result

18

• With speed disturbance 0.2m/s magnitude of WGN

Page 19: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Summary

19

Guiding Attraction based Random Tree(GART)Six-order Bezier curve and Tuning RotationForward simulation to find dangerous regionLocal Optimal Reshaping

Page 20: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

Limitation

• No mention about computational time

• No dynamics

• No difficult obstacles

20

Page 21: Generation of Dynamically Feasible and Collision Free ...sungeui/MPA_S17/Presentation/...• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order

[CS686] Paper presentation 1

KAIST School of Computing

QnA

21

THANK YOU

References• Generation of Dynamically Feasible and Collision Free Trajectory by Applying Six-order Bezier Curve and Local Optimal Reshaping (IROS 2015)• Guiding attraction based random tree path planning under uncertainty: Dedicate for UAV (IROS 2014)