Vous êtes sur la page 1sur 31

Cooperative Control of Multiple Robotic Vehicles:

Theory and Practice

Trajectory tracking & Path-following control


EECI Graduate School on Control
Suplec, Feb. 21-25, 2011

A word about T Tracking and P Following


Path following
Reference path given in a
time-free parameterization
Constant forward velocity
Smoother convergence to
the path
Reference path

Space

Trajectory Tracking
Time and space reference
trajectory
The vehicle may turn back
in its attempt to be at a
given reference point at a
prescribed time
Reference
trajectory

Possible
vehicle
trajectory

Space x Time

Path-following is motivated by applications in which spatial errors are more critical than
temporal errors

Reference-tracking versus path-following

Additional design of freedom

The reference-tracking problem is subjected to the limitations


imposed by the unstable zero-dynamics.
The path-following problem is not subjected to these limitations
The freedom to design a timing law is a major advantage of pathfollowing over reference tracking.
A. Pedro Aguiar, Joo P. Hespanha, and Petar Kokotovi, Path-Following for Non-Minimum Phase
Systems Removes Performance Limitations.
IEEE Transactions on Automatic Control, Vol. 50, No. 2, pp. 234-239, Feb. 2005.

Position tracking of an underactuated Hovercraft


Goal: force the hovercraft to track a
circular trajectory (black)

due to side-slip the velocity of the hovercraft is not tangent to the trajectory

Problem statement

Consider an underactuated vehicle modeled as a rigid body subject to external


forces and torques

Kinematics

Dynamics

Trajectory-tracking problem

Given a trajectory pd:[0,1) ! R3, we want the tracking error p(t)-pd(t)


to converge to a neighborhood of the origin that can be made
arbitrarily small
The solution should be robust with respect to
parametric modeling uncertainty
later extended to path following

Position tracking of an
underactuated Hovercraft
Goal: force the hovercraft to track a
circular trajectory (black)

due to side-slip the velocity of the hovercraft is not tangent to the trajectory

Controller design

Step 1. Coordinate Transformation


tracking error in
body frame

Step 2. Convergence of e
error only in position!

<0

linear velocity viewed as


a virtual control input

Controller design

virtual control input

Step 3. Backstepping for z1

control input

It will not always be possible to drive z1 to zero!


Instead, we will drive z1 to a small constant

<0
dominated by the first term

<0
1st control signal has
been assigned
angular velocity viewed
as a virtual control input

Controller design

Step 4. Backstepping for z2

<0
dominated by the first term

<0

<0
2nd control signal
has been assigned

Using Youngs inequality, 9 > 0


can be made
arbitrarily small

All signals remain bounded and converges to ball of radius proportional to

Position tracking of an
underactuated Hovercraft
Goal: force the hovercraft to track a
circular trajectory (black)

due to side-slip the velocity of the hovercraft is not tangent to the trajectory

Model parameter uncertainty

What happens if there is parametric modeling uncertainty?

Coefficient of viscous friction assumed used by the controller is 10% of the real value

Closed-loop system still stable but considerable performance degradation

Supervisory control
supervisor

controller 1
bank of candidate
controllers

controller n

exogenous
disturbance/
noise

switching signal

process

control
signal

Key ideas:
1. Build a bank of alternative controllers
(one for each possible value/range of the unknown parameter)
2. Supervisor places in the feedback loop the controller that seems
more promising based on the available measurements

measured
output

Estimator-based supervisions setup


w
control
signal

exogenous
disturbance/noise

process

measured
output

Process is assumed to be in a family

parametric uncertainty

For each admissible process model Mp,


there is one candidate controller Cp that provides adequate performance.
How to determine which admissible model matches the real process?

Estimator-based supervisor
measured
output
control
signal

y
y
u

multiestimator

decision
logic

switching
signal

Process is assumed to be in family


process is
Mp, p2 P

controller Cp provides
adequate performance

Multi-estimator
yp estimate of the process output y that would be correct if the process was Mp
ep output estimation error that would be small if the process was Mp
Decision logic:
ep small

process is
likely to be Mp
Certainty equivalence inspired

should
use Cp

Multi-estimator for the vehicle model

Process model

uncertainty in the dynamic equations through: M, J, fv , f

Family of estimator equations (p2 P)

scalar positive functions

estimation errors

The correct estimation error ep* satisfies


convergence to zero
small integral norm

Estimator-based supervisor
measured
output
control
signal

y
y

multiestimator

decision
logic

switching
signal

+ A stability argument cannot be


based on this because typically
Process is assumed to be in family
process is Mp ) ep small
process is
controller Cp provides
but not the converse
Mp, p2 P
adequate performance
Multi-estimator
yp estimate of the process output y that would be correct if the process was Mp
ep output estimation error that would be small if the process was Mp
Decision logic:
ep small

process is
likely to be Mp

use Cp

Certainty equivalence inspired

Estimator-based supervisor
measured
output
control
signal

y
y

multiestimator

decision
logic

switching
signal

Process is assumed to be in family


process is
Mp, p2 P

controller Cp provides
adequate performance

Multi-estimator
yp estimate of the process output y that would be correct if the process was Mp
ep output estimation error that would be small if the process was Mp
detectable means
small ep ) small state

Decision logic:
ep small

process is
likely to be Mp

use Cp

overall system
is detectable
through ep

overall state
is small

Certainty equivalence inspired, but formally justified by detectability

Detectability property
detectable means
small ep ) small state

ep small

process is
likely to be Mp

use Cp

overall system
is detectable
through ep

Using the original Lyapunov function and


positive
constant

in L1

constant

in L2

When ep is small, all signals remain bounded

in L1

& e converges to ball of radius proportional to

overall state
is small

Scale-independent hysteresis switching


How to pick a small estimation error ?
monitoring signals integral norm of estimation error

start

p2P

hysteresis
constant

measure of the size of ep over a


window of length 1/

y
n
wait until current monitoring
signal becomes significantly
larger than some other one

forgetting factor

Simulation results
Estimator-based supervisor controller
0.5

ex [m]

-0.5
-1

10

20

30

40

50

60

40

50

60

40

50

60

40

50

60

time [s]

0.5

ey [m]

-0.5
-1

10

20

q-q [degree]
d

30

time [s]

50

-50

10

20

30

time [s]
8

dv

10

20

30

time [s]

Position tracking in the presence of parametric uncertainty and measurement noise

Path-following

Consider an underactuated vehicle modeled as a rigid body subject to


external forces and torques

Kinematics

Dynamics

Path-following problem

Given a geometric path {pd()2 R3 : 2[0,1)} and speed assignment vr()


2 R, we want

the position to converge and remain inside a tube centered around the
desired path than can be made arbitrarily thin, and
satisfy (asymptotically) the desired speed assignment, i.e., ! vr as
t!1

Tracking and path-following of


an underwater vehicle (3-D space)

The Sirene AUV developed


for Deep Sea Intervention on
Future Benthic Laboratories

Goal: force the underactuated


AUV to track a desired
helix trajectory

Path-following

Goal

Given a geometric path {pd()2 R3 : 2[0,1)} and speed assignment vr() 2 R, we want

the position to converge and remain inside a tube centered around the desired path
than can be made arbitrarily thin, and
satisfy (asymptotically) the desired speed assignment, i.e., ! vr as t!1

Define
speed error

Choosing

Same conclusions as before

Simulation results
Trajectory tracking

Path-following

roll [degree]

50

100

200

250

300

time [s]

-20

50

100

150

200

250

300

time [s]

200

yaw [degree]

150

50

100

150

time [s]

-5

50

100

200

250

300

150

200

250

300

200

250

300

200

250

300

time [s]

20

-20

50

100

150

time [s]

200

-200

pitch [degree]

-5
20

yaw [degree]

pitch [degree]

roll [degree]

-200

50

100

150

time [s]

UAV Path Following Concept

Path-following for an Unmanned


Aerial vehicle (UAV)

Objective: follow predefined spatial 3D paths


paths are time-independent:
decoupling between space and time = separation of 3D path and speed
speed can be used as an additional DOF for time coordination

UAV Path Following


System Architecture
Inner/Outer Loop Solution

L1 adaptive
controller
Trajectory
Generation

polynomial
path

Path following
(Outer loop)

Onboard PC104
Boundary
conditions

User Laptop

Pitch rate
Yaw rate
commands

Onboard A/P
(Inner loop)

Problem Geometry
F: Serret-Frenet frame
W: wind frame
I: inertial frame

UAV
Q
Inertial
Frame {I}

desired trajectory
zI
UAV speed
flight path angle
heading angle
Desired path
path length
to follow
position
of UAV in inertial frame

3D Kinematics Equations

Vs
zF(B)

yF(N)
y1

qI
yI

qF

z1

xF(T)

s1
P Serret Frenet
Frame {F}

pc
xI

Input:

UAV Path Following


Key idea: use virtual target to determine desired location on the path
Minimize the distance of the UAV from the virtual target on the path
Reduce the angle between the vehicle velocity vector and local tangent to the path

Virtual targets motion extra degree of freedom


Path

Q
q

s1

y1
{I} : Inertial Frame

Virtual Target

{F} : Serret-Frenet Frame

UAV Path Following (cont.)


Control the evolution of the virtual target : added degree of freedom

Path

{I} : Inertial Frame

Kinematics
F

Coordinat
e
systems

- difference between

Kinematics equations in I

where

and

- Euler angles from F to W

Error Equations in F

in F

Kinematic Control Law

Desired shaping functions

Path following control laws

Vous aimerez peut-être aussi