hasan arshad nasir - cyphynets · rrt algorithm for generating large set of candidate paths....

30
Hasan Arshad Nasir

Upload: others

Post on 06-Aug-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 2: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Motion Uncertainty.

Imperfect state information.

Page 3: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 4: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 5: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 6: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Quantify these uncertainties in advance.. Develop optimized path planning..

Page 7: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

LQG-MP (Linear Quadratic Gaussian- Motion Planning).

Page 8: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 9: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

RRT algorithm for generating large set of candidate paths.

Generating and Linearizing Dynamics and Observation model.

Kalman filter for optimal state estimation and LQR for optimal control.

Probability distribution of state and control input.

Based on above choose the best path.

Page 10: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

RRT algorithm for generating large set of candidate paths…………….

Page 11: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Its “Rapidly Exploring Random Tree”. It’s the best for this context. Its probabilistic complete algorithm.

Page 12: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

1) http://www.kuffner.org/james/plan/algorithm.php

Page 13: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 14: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 15: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 16: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Generating and LinearizingDynamics and Observation model.…………….

Page 17: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

).,0(~).,0(~

).()()()().,(

).()()()()()1().,,( 11

tt

tt

ttt

tttt

NNnMNm

knkxkHkznxhz

kmkukBkxkAkxmuxfx

+==

++=+= −−

Page 18: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

.)(

)()(

.)()(

)(

.)1()1(

)1(

=

=

++

=+

ky

kxku

kykx

kx

kykx

kx

r

r

r

r

r

r

.1

.1001

)(

.0

0)(

.1001

)(

=

=

=

=

t

kH

tt

kB

kA

Page 19: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

.)()(

)()(

1001

)()(

.)()(

)(

)(0

0)()(

1001

)1()1(

+

=

+

+

=

++

knkn

kykx

kzkz

kmkm

ky

kxt

tkykx

kykx

y

x

r

r

y

x

y

x

r

r

r

r

r

r

Page 20: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Kalman filter for optimal state estimation and LQR for optimal control..…………….

Page 21: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Process update (prediction). Measurement update (update/filtering).

Page 22: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Cost function:

“lqr” command in MATLAB, generates feedback matrix.

.))()()()((0∑=

+l

kr

Trr

Tr kuRkukxQkx

).()( kLxku rr =

Page 23: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Red star = estimated stateGreen circle = true stateMagenta star = measured state

Page 24: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Probability distribution of state and control input........

Page 25: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 26: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Designing a planning objective and selecting the best path........

Page 27: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state
Page 28: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Path serial number Color Weight/Ratio/Priority

1 Red 2.769

2 Blue 2.8095

3 Black 2.872

4 Yellow 2.880

5 Green 2.795

Page 29: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Jur V.Berg, P. Abbeel, K. Goldberg, “LQG-MP: Optimized Path Planning for Robots with Motion Uncertainty and Imperfect State Information”. International Journal of Robotics, 2011.

Page 30: Hasan Arshad Nasir - CYPHYNETS · RRT algorithm for generating large set of candidate paths. Generating and Linearizing Dynamics and Observation model. Kalman filter for optimal state

Dr. Abubakr Muhammad. Talha Manzoor.

[email protected]