Passivity Based Control

Eolss Publishers Co. Ltd.
,
Oxford, United Kingdom
Copyright © 2009 EOLSS Publishers/ UNESCO
Rev 1.1, October 2009
Information on this title: www.eolss.net/eBooks
ISBN 978-1-84826-152-5 (e-Book Adobe Reader)

ISBN 978-1-84826-602-5 (Print (Full Color Edition))
The choice and the presentation of the facts contained in this publication and the opinions expressed
therein are not necessarily those of UNESCO and do not commit the Organization.
The designations employed and the presentation of material throughout this publication do not imply the
expression of any opinion whatsoever on the part of UNESCO concerning the legal status of any country,
territory, city, or area, or of its authorities, or the delimitation of its frontiers or boundaries,
This information, ideas, and opinions presented in this publication are those of the Authors and do not
represent those of UNESCO and Eolss Publishers.
Whilst the information in this publication is believed to be true and accurate at the time of publication,
neither UNESCO nor Eolss Publishers can accept any legal responsibility or liability to any person or
entity with respect to any loss or damage arising from the information contained in this publication.
All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopying, recording, or any information storage or
retrieval system, without prior permission in writing from Eolss Publishers or UNESCO.
The above notice should not infringe on a 'fair use' of any copyrighted material as provided for in
section 107 of the US Copyright Law, for the sake of making such material available in our efforts to
advance understanding of environmental, political, human rights, economic, democracy, scientific, and
social justice issues, etc. If you wish to use copyrighted material from this e-book for purposes of your
own that go beyond 'fair use', you must obtain permission from the EOLSS Publishers.
Every effort has been made to trace and credit all the copyright holders, but if any have been
inadvertently overlooked, UNESCO and Eolss Publishers will be pleased to make the necessary
arrangements at the first opportunity.
British Library Cataloguing-in-Publication Data

A catalogue record of this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data

A catalog record of this publication is available from the library of Congress
Singapore
CONTROL SYSTEMS, ROBOTICS AND AUTOMATION - Vol. XIII - Passivity Based Control - Antonio Loría and Henk
Nijmeijer
PASSIVITY BASED CONTROL

Antonio Loría
CNRS, LSS-Supélec, Plateau de Moulon, 91192, Gif sur Yvette, France
Henk Nijmeijer
Departments of Mech. Engg., Eindhoven Univ. of Technology, The Netherlands
Keywords: Passivity, automatic control, mechanical systems, robot manipulators,

Euler-Lagrange systems, nonlinear systems, input-output stability.
Contents
1. Introduction
2. Passivity: mathematically speaking
3. Stability of passive systems
4. PBC of Euler-Lagrange systems
5. Epilogue
Glossary
Bibliography
Biographical Sketches
Summary
Passivity is a fundamental property of many physical systems which may be roughly

defined in terms of energy dissipation and transformation. It is an inherent Input-Output
property in the sense that it quantifies and qualifies the energy balance of a system when
stimulated by external inputs to generate some output. Passivity is therefore related to
the property of stability in an input-output sense, that is, we say that the system is stable
if bounded “input energy” supplied to the system, yields bounded output energy. This is
in contrast to Lyapunov stability which concerns the internal stability of a system, that
is, how “far” the state of a system is from a desired value. In other words, how
differently a system behaves with respect to a desired performance.
Passivity based control is a methodology which consists in controlling a system with the
aim at making the closed loop system, passive. The field constitutes an active research
direction and therefore in this chapter we give only a basic overlook of the most
important concepts involved. A section is also devoted to a wide class of physical
passive systems: the Euler-Lagrange (EL) systems and their passivity-based control.
The reader should rather consider this presentation as very concise image of the material
cited in the Bibliography. Therefore, we invite the reader who wishes to obtain a deeper
knowledge in the subject, to see those references.
1. Introduction
To better understand the passivity concept and passivity-based control (PBC), we need
to leave behind the notion of state of a system and think of the latter as a device which
©Encyclopedia of Life Support Systems (EOLSS) 206

Nijmeijer
interacts with its environment by transforming inputs into outputs. From an energetic
viewpoint we can define a passive system as a system which cannot store more energy
than is supplied by some “source”, with the difference between stored energy and
supplied energy, being the dissipated energy.
Hence, it shall be clear that passivity is closely related to the stability of a system, in the
input-output sense evoked in the Summary. In PBC achieving stability from this
viewpoint is the first goal.
A fundamental property of passive systems is that, regarding a feedback interconnection

of (other physical) passive systems, passivity is invariant under negative feedback
interconnection. In other words, the feedback interconnection of two passive systems
yields a passive system.
Thus, if the overall energy balance is positive, in the sense that the energy generated by
one subsystem, is dissipated by the other one, the closed loop will be stable in an input-
output sense (see Proposition 2). This property constitutes the basis of passivity-based
control (PBC).
The term PBC was coined in 1989 in the context of adaptive control of robot
manipulators to define a controller methodology whose aim is to render the closed-loop
system passive, seen as a map from an external new input. This objective seemed very
natural within that context, since the robot dynamics defines a passive map from input
torques to output link velocities. As a matter of fact this passivity property is inherent to
many other physical systems such as electrical and electromechanical. (See section 4).
Since the aim in PBC is to render the closed loop system passive, the main property
used in PBC is the fact that the interconnection of passive systems is passive.
Conversely, passive systems can be decomposed in passive “subsystems”. Thus, in this
philosophy the controller may be designed as a passive system.
In terms of energy dissipation, the PBC approach may be viewed as an extension of the
so-called energy-shaping plus damping injection technique introduced to solve state-
feedback set (operating) point regulation problems in fully actuated robotic systems
back in 1981. For this particular problem we can concentrate our attention on the
potential energy and the dissipation functions to proceed along two basic stages: firstly,
as energy shaping stage which consists on modifying the potential energy of the system
in such a way that the “new” potential energy function has a global and unique
minimum at the desired equilibrium. This is motivated by the well known fact (stated by
Joseph Lagrange in 1788 and proved 50 years later by Dirichlet) that the stable
equilibria of mechanical systems correspond to the minima of the potential energy
function. Secondly, a damping injection stage which consists in modifying the
dissipation properties of the system, to render it strictly passive.
Viewed from the PBC perspective the energy shaping stage accomplishes the objective
of rendering the closed loop system passive with a desired storage function that consists
of the original kinetic energy and the new desired potential energy. The damping
injection reinforces this property to output strict passivity. Finally, Lyapunov

Nijmeijer
asymptotic stability follows from the input-output stability of the output strictly passive
map provided some dissipation propagation (i.e., detectability) conditions are met. That
is, the system evolves in a way that it reaches the desired set point asymptotically. See
Sections 2.1 and 3.2.
The generality of the PBC allows us to deal with different problems such as output
feedback and tracking control in a unified way. Moreover, even though here we will
only illustrate the PBC methodology with simple examples of control of EL mechanical
systems, the reader must keep present that, having its origins in electrical circuits, it is
natural that PBC is most suitable for electrical and electromechanical systems such as
power converters, electrical machines, etc. This will be illustrated through a time-
varying reference tracking control problem in Section 4.4.
2. Passivity: Mathematically Speaking
In this section we will introduce the precise definitions of passivity and some important
theorems on passivity.
2.1. In a General Input-Output Framework
As it may be clear from the discussion above, when talking about a passive system
(operator) one aims at measuring the energy (storage) transformation performed in the
system. The concept of passivity in dynamical systems has its roots in the same concept
used by electrical engineers to characterize elements which consume energy but do not
supply it. In this context, the input and output signals have a direct physical meaning,
i.e., current and voltage hence, the (electrical) energy measure of these is evident: it
simply corresponds to the integral of the power over time. However, if we would like to
talk about the passivity property of physical systems of different nature (electrical,
mechanical, chemical, etc), we need a more general concept of measure.
To that end, we must keep in mind that passivity is a property of the system, seen as an
operator which maps inputs into outputs. In this respect, we will find characterizations
and sufficient conditions for passivity, which apply to systems that can be modeled by
rational transfer functions as well as to systems modeled by nonlinear (possibly time-
varying) differential equations.
In this section we will introduce the precise definitions of passivity, which reflect the
fact that passivity is an energy transformation property. We will also extend to the case
of nonlinear systems, some of the arguments made before, to sustain the fundamental
properties of passive systems.
Definition 1 ( L2 and L2 norms): The L2 norm of a signal f : \ ≥ 0 → \ , f (t ) is

n n n n n
denoted f (t ) 2T and defined by

Nijmeijer
(∫ )
1
T 2 2
0
f (t ) dt (1)
n
and the L 2 norm denoted f (t ) 2 is defined as by lim f (t ) 2T .
T →∞
With this metric we can then define the normed L2e -space:
Definition 2 ( L2e -space): We say that f : \ ≥ 0 → \ belongs to L2e if and only if

f (t ) 2T
<∞
The definitions above makes sense from a practical viewpoint if we consider the case
n
when f (t ) corresponds to power and therefore, the L2 borrows the interpretation of
energy amount over a time interval.
n
Now in order to properly define the passivity concept for L2 signals we introduce the
following product, which generalizes the concept of supplied energy discussed above.
Definition 3 (Inner product): Let u , y ∈ L2 and T > 0 , then the inner product is
n
defined ∀T > 0 by
T
u| y T
:= ∫0 u (t ) y (t ) dt . (2)
For illustration we may consider u to be an input voltage supplied to an electrical load

in which case i may be interpreted as the output current. Therefore, inner product gives
a measure on the energy stored in the resistor.
With these tools we can now properly define the following
Definition 4 (Passivity): An operator H : u 6 y is passive if there exists a β ∈ \ such

that
u| y T
≥ β. (3)
The number β depends on the initial conditions of the signals. Often, it quantifies the
initial energy stored in the system. This will become clearer when dealing with passivity
of mechanical systems, in section 4.1.
Definition 5 (Output Strict Passivity): An operator H : u 6 y is output strictly

passive if there exists β ∈ \ and δ o > 0 such that

Nijmeijer
2
u| y T
≥ δo y 2T
+ β. (4)
Definition 6 (Input Strict Passivity): An operator H : u 6 y is input strictly passive

if there exists β ∈ \ and δ i > 0 such that
u| y T
≥ δi u 2T +
2
β. (5)
Passive interconnected systems
The following theorems formalize the fact that passivity is sustained for the
interconnection of passive systems.
Theorem 1: Consider the input-output system depicted in Figure 1. Let

e := (e1 , e2 ), u := (u1 , u2 ) and y := ( y1 , y2 ) be in L22en . If ∑1 and ∑ 2 are both passive
then ∑ : u 6 y is also passive. If ∑1 and ∑ 2 are OSP then ∑ : u 6 y is also OSP.
Figure 1: Feedback interconnection of passive systems
The theorem below regards a special case of the feedback interconnection depicted in
Figure 1, when u2 ≡ 0 . This structure is particularly important since it is the typical
case of a plant (∑1 ) in closed loop with a controller (∑ 2 ) . In this case the input
u1 plays the role of an external signal to the closed loop. Notice that this input can be in
its turn the output of another passive block. In this way one can build a new passive
system upon a core passive block. Therefore these theorems are fundamental to
passivity-based control.
Theorem 2: Consider the closed loop system of Figure 1, with u2 ≡ 0 .Assume that
∑i : L2ne 6 L2ne , i = 1,2 . Then e2 = y1 ∈ L2ne if either of the following statements is
true:

Nijmeijer
• If ∑1 : e1 6 y1 is passive and ∑ 2 : y1 6 y2 is ISP or

• If ∑1 : e1 6 y1 is OSP and ∑ 2 : y1 6 y2 is passive.
So far we have stated formally and in fair generality, under which conditions a feedback
interconnection of passive systems yields a passive system. However, it will be also
useful to know that interconnections not only preserve the passivity properties of the
subsystems but, in certain cases, passivity can be strengthened. To illustrate this idea,
we briefly discuss next, a technique called loop transformation.
Consider the interconnected system of Figure 1 with only one input, i.e., let u2 ≡ 0 .
Assume that ∑ 2 is ISP and ∑1 is passive. The loop transformation technique will make
evident that since the system ∑ 2 is “more dissipative” than ∑1 (some readers will know
that the term “dissipative” and “passive” are mathematically different. With an abuse of
notation we use here the term dissipative to denote a system which dissipates energy in
a non recoverable manner, e.g., heat.), by performing the interconnection ‘some’ of the
“dissipation of ∑ 2 is propagated to ∑1 ”. To show this we will use Figure 2, which
represents a system equivalent to that of Figure1 with u2 ≡ 0 , and the following
Fact 1: Assume that the system ∑ 2 is ISP and has finite L2 gain, i.e., there exists
∞ > c > 0 such that y2 2T ≤ c e1 2T . Then, the map ∑ 2 is also OSP.
Let us perform a few simple calculations to exhibit the new passivity properties of the
feedback interconnected system of Figure 2. For ∑1′ : e1 6 y1 , using the passivity
property of ∑1 , we have that
e1 | y1 2T
= u1 − y2 + ky1 | y1 = u1 − y2 | y1 + ky1 | y1 2T
2T 2T
≥ β1 + k y1 2T
.

Nijmeijer
Figure 2: Loop transformed feedback interconnected system
That is, the loop transformation has rendered the map ∑1′ , OSP. The price paid for this
is that the ISP of ∑ 2 has been “weakened” more precisely,
y1 | y2 − ky1 2T
= y1 | y2 2T
− ky1 | y1 2T
≥ β 2 + δ i 2 y1 2T
− k y1 2T
hence, we must impose k < δ i 2 .
It is important to remark at this point that the coefficient k is used only for analysis
hence, there is no loss of generality in restricting it to be k < δ i 2 . Notice also that the
physical system has not changed with the loop transformation but only the way we look
at it!
Using the Fact 1 we obtain that the system of Figure 1 with ∑1 passive and ∑ 2 ISP and
finite L2 gain is equivalent to the interconnection of an OSP with an OSP and ISP
system. This observation is sometimes fundamental in the stability analysis of passive
systems, and consequently in passivity –based control, as we will see in section 3.2 and
4.2.
3. Stability of Passive Systems
Before discussing PBC we need to discuss about stability. In particular, the type of
stability which one pursues in PBC, is, in an input-output sense.

Nijmeijer
3.1. L2 -Stability
A “relaxed” definition of L2 stability is that an operator is L2 stable if it maps

L2 inputs into L2 outputs. However, in a more strict sense one may also be interested in
measuring the “amount of stability”. This is given by the so-called L2 gain.
Definition 7 ( L2 -stability): The state space system ∑ is said to be L2 stable with

finite L2 gain if there exists a positive constant γ such that for every initial condition
x0 = x (0) , there exists a finite constant β( x0 ) such that
y (t ) 2T
≤ γ u (t ) 2T
+ β( x0 ) .
Proposition 1: If ∑ : u 6 y is OSP then it has finite L2 -gain.
Proof. The proof follows straight forward observing that OSP implies the existence of
2
2
δ > 0 and β ∈ \ such that δ o y ≤ u| y −β+ 1 1 u − δo y , therefore
2T T 2 δo
2T
δo 2 2
2
y 2T
≤ 21δ u 2T
− β . Thus the L2 gain γ ≤ δ1 .
o o
3.2. From L2 -Stability to Lyapunov Stability
Fundamental concepts that relate the input-output stability with the stability in the sense
of Lyapunov are zero-state detectability and zero-state observability.
Definition 8 (Zero-state detectability): A state-space system

x = f ( x ), x ∈ \ , f (0) = 0 is
n
zero-state detectable from the output
y = h( x), h(0) = 0 , if for all initial conditions x(0) ∈ \ n we have
y (t ) ≡ 0 ⇒ lim x(t ) = 0 .
t →∞
Definition 9 (Zero-state observability): A state-space system

x = f ( x ), x ∈ \ , f (0) = 0 is
n
zero-state observable from the output
y = h( x), h(0) = 0 if for all initial conditions x(0) ∈ \ n we have
y (t ) ≡ 0 ⇒ x(t ) ≡ 0 .
Roughly speaking, thinking of physical systems these properties can be regarded as

energy dissipation propagation properties. They establish that the asymptotic
convergence of the state x of a system will follow from that of the output y = h( x ) if
the latter is “suitably defined”. In general these properties may be difficult to verify and

Nijmeijer
in occasions impossible with respect to some outputs. The lemmas below are
particularly useful to establish a link between the asymptotic convergence of an output
h( x) and the state.
Lemma 1: Let y = G ( p )u , where G ( p ) is an n × m strictly proper, stable transfer

function and p= d
dt
. Then y ∈ L2n implies that u ∈ L2n ∩ L∞n , u ∈ L2n , y (t ) is
continuous, and y (t ) → 0 as t → ∞ . If in addition, u (t ) → 0 as t → ∞ ,
then y (t ) → 0 .
Lemma 2: If the system ∑ : u 6 y is OSP, y ∈ L2 ∩ L∞ , y (t ) ∈ L∞ and, the system

n n n
is zero state detectable from this output then, if u ≡ 0 the state trajectories x(t ) → 0 as
t → ∞.
We finish this section with a proposition which summarizes the results recalled here and
which are fundamental in passivity-based control of EL systems. Assume that the
systems
⎧⎪ x i = fi ( x i ) + gi ( x i )ui
∑i ⎨ , i = 1, 2, fi (0) = 0, hi (0) = 0
⎪⎩ yi = hi ( x i )
are passive or output strictly passive, with storage functions H1 ( x1 ) and

H 2 ( x2 ) respectively, i.e.
t 2 t
H1 ( x1 (t ) − H1 ( x1 (0)) + ∫ 0 y1 ( s ) ≤ ∫ 0 u1 ( s ) y1 ( s )ds (6)
t 2 t
H 2 ( x2 (t ) − H 2 ( x2 (0)) + ∫ 0 y2 ( s ) ≤ ∫ 0 u2 ( s ) y2 ( s ) ds , (7)
then the following holds true.
Proposition 2:
i. (i) Suppose ∑1 and ∑ 2 are passive (respectively output strictly passive), then the
feedback interconnected system ( ∑1 , ∑ 2 ) of Figure 1, defines a passive
(respectively output strictly passive) operator (e1 , e2 ) 6 ( y1 , y2 ) .
∗
ii. (ii) Suppose that H1 and H 2 , satisfying (6)-(7) have strict local minima in x1 ,
∗ ∗ ∗
respectively x2 , then ( x1 x2 ) is a stable equilibrium of the feedback system,
with e1 = e2 = 0 .

Nijmeijer
iii. (iii) Suppose that ∑1 and ∑ 2 are output strictly passive and zero-state
detectable, and that H1 and H 2 , satisfying (6)-(7), and proper, have a global
∗ ∗
and unique minimum in x1 = 0 , respectively x2 = 0 then (0,0) is a globally
asymptotically stable equilibrium of the feedback system ( ∑1 , ∑ 2 ) with
e1 = e2 = 0 .
4. PBC of Euler-Lagrange Systems
In this section we will study a class of passive systems which includes electrical,
mechanical and electromechanical systems. The systems we will study borrow their
name to the equations which we use to model their dynamics: the Euler-Lagrange
equations. These equations can be derived either using laws of force and the so-called
D’Alembert’s principle (see any of the texts on Mechanics cited in the Bibliography) or
by means of the calculus of variations. The starting point of the variational approach is
to define the energy functions in terms of sets of generalized coordinates (these are
coordinates of different nature, e.g., they can be the charges q for an RLC circuit or the
position angles of a motor shaft or of a pendulum) and then, the introduction of the so
called Lagrangian function which is the difference between the kinetic and the potential
energy. The equations of motion are then derived invoking well-known principles of
analytical dynamics, for instance the fundamental Hamilton principle, which roughly
speaking states that the system moves along trajectories that minimize the integral of the
Lagrangian along a trajectory q (t ) .
We will study PBC of EL systems because they are passive systems and therefore, PBC
is most suitable for them. Moreover PBC underscores the role of the interconnections
between the systems and provides us with the storage and dissipation functions.
Without further introduction let us consider a dynamical system with n degrees of

freedom, with generalized coordinates q ∈ \ and external forces Q ∈ \ . Such a
n n
system is an Euler-Lagrange system if its dynamics is described by the equations
d ⎛ ∂L ⎞ ∂L
⎜ (q, q ) ⎟ − (q, q ) = Q , (8)
dt ⎝ ∂q ⎠ ∂q
where
L( q, q ) := T ( q, q ) − V ( q ) (9)
is the Lagrangian function, T ( q, q ) is the kinetic energy (or co-energy) function which
we assume to be of the form

Nijmeijer
1 T
T (q, q ) = q D (q )q , (10)
2
where D ( q ) ∈ \ n × n is the generalized inertia matrix that satisfies

D (q ) = D T ( q ) > 0 , and V (q) is the potential function which is assumed to be
bounded from below that is, there exists a c ∈ \ such that V ( q ) ≥ c for all q ∈ \ .
n
In a fairly general context, we may consider three types of external forces: the action of
controls, dissipation and the interaction of the system with its environment. We will
n× n u
assume controls to enter linearly as Mu ∈ \ , where M ∈ \
n
is a constant matrix
is the control vector. Dissipative forces are of the form − ∂∂Fq ( q ) , where
nu
and u ∈ \
F (q ) is the Rayleigh dissipation function which by definition satisfies
∂F
q (q ) ≥ 0 . (11)
∂q
In summary the external forces can be expressed by
∂F
Q=− (q ) + Qς + Mu , (12)
∂q
where Qς is an external signal that models the effect of disturbances.
As explained in the Introduction, in PBC the control objective is achieved by imposing

to the closed loop dynamics a certain passivity property, which in its turn reduces to
assigning some desired storage and dissipation functions. It comes natural then to define
EL systems in the following manner
Definition 10 (EL equations and EL parameters): The EL equations of motion
d ⎛ ∂L ⎞ ∂L ∂F
⎜ (q, q ) ⎟ − (q, q ) + (q ) = Mu + Qς , (13)
dt ⎝ ∂q ⎠ ∂q ∂q
with (9), (10), (11) define an EL system which is characterized by its EL parameters:
{T (q, q ), V (q ), F (q ), M, Qς } .
The use of the EL parameters as defined by the quintuple above captures a fairly general
notation. However, when clear from the context, we may use the more compact notation
{T ( q, q ), V ( q ), F ( q ), M} for systems for which Qς ≡ 0 or

Nijmeijer
{T ( q, q ), V (q ), F ( q )} in the case when only dissipative forces affect the EL system.
The matrix M is a full column rank matrix relating the external inputs to the
generalized coordinates. We find it convenient to distinguish two classes of EL systems
according to the structure of this matrix:
Definition 11 (Underactuated EL systems.): An EL system is fully-actuated if it has

equal number of degrees of freedom than available control inputs (that is if n = nu , e.g.
if M = I n ). Otherwise, if n u < n we say that the system is underactuated. In the latter
⊥
case, q can be partitioned into non-actuated M q and actuated components Mq ,
⊥
where M denotes the perpendicular complement of M .
A second classification that we find convenient to introduce at this point, involves the
presence of damping. We can thus distinguish two classes of systems:
Definition 12 (Underdamped and fully-damped systems.): The EL system (13) is

said to be fully-damped if the Rayleigh dissipation function satisfies
∂F n
q (q ) ≥ ∑ α i qi2
∂q i =1
with α i > 0 for all i ∈ n := {1,..., n} . It is, on the other hand, underdamped if
∃ i ∈ n such that α i = 0 .
4.1 Passivity of EL Systems
Before discussing the method of passivity based control, we will show that EL systems
are passive. More precisely, they define passive maps Q 6 q .
Proposition 3: The EL system (13) with Qς ≡ 0 defines a passive operator

Σ : u 6 M q with storage function the system’s total energy H(q, q ) . That is,
u | M q ≥ H[q(T ), q (T )] − H[q(0), q (0)] (14)

T
for all T ≥ 0 and all u ∈ L2e . Further, this property is strengthened to output strict
m
passivity (OSP) if the system is fully damped. In this case
2
u | M q ≥ α M q + H[q (T ), q (T )] − H[q (0), q (0)] (15)
T 2T

Nijmeijer
for some α > 0 and all u ∈ L2e .

m
The inequality (15) comes from the fact that (damped) EL systems are (strictly) passive.
This can be seen from the energy balance equation
T ∂F (q) T
H[q(T ), q(T )] − H[q(0), q(0)] + ∫0 q ds = ∫0 q Muds , (16)

∂q
stored energy
supplied
dissipated
which EL systems satisfy. Observe that, since V (q ) is bounded from below by c , and
T ( q, q ) ≥ 0 we have that H(q, q ) ≥ c . If furthermore the Rayleigh dissipation
function satisfies (11) hence (14) follows. If the system is fully damped it follows
min i {α i }
immediately from Definition 12 and (16) that (15) holds with α := 2
.
M
The energy balance equation (16) reveals several interesting properties of EL systems:
1. If we set u = 0 we see that energy is not increasing, hence the trivial equilibrium
of the unforced system is stable in the sense of Lyapunov. As a matter of fact
these considerations constituted the starting point of A.M. Lyapunov’s original
work.
2. Stability is also preserved if we now fix the output Mq to zero, hence reflecting
the fact that the system is minimum phase (i.e. the dynamics compatible with the
constraint Mq ≡ 0 is Lyapunov stable).
3. We see that damping can be easily added-along the actuated channels- if q is

measurable. Notice, however, that the operator u 6 M q may be output
strictly passive even if energy is not dissipated “in all directions”. Namely, it is
∂F ( q ) 2
enough to ensure q ∂q
≥ α M q . This property will be used in the
sequel.
4.2. PBC of EL Systems
4.2.1. An Introductory Example
The passivity-based technique we will discuss is known also as energy shaping plus
damping injection. To that end, let us start with the expression of the total (kinetic +
potential) energy of the simple pendulum,

Nijmeijer
1
Hp (q, q ) = ml 2 q 2 + mgl (1 − cos(q )) ,
2

V( q )
T ( q , q )
where q ∈ \ and g is the gravity acceleration. We assume torque as the control input
u , hence in the absence of friction the EL parameters of such system are
{T ( q, q ), V ( q ),0,1} . Using the EL equations we can easily derive the dynamics
ml 2 q + g ( q ) = u , (17)
where g (q ) , the gravitational force, is the force derived from the potential energy, that
is,
∂V (q)
g (q) := = mgl sin(q) .
∂q
Our design problem is to stabilize the pendulum at a constant

equilibrium [ q, q ] = [ q∗ ,0] . According to the energy-shaping plus damping injection
procedure, we will seek to modify the potential energy and the Rayleigh dissipation
function of the system, leaving untouched the kinetic energy, since it plays no role on
the stability properties of the equilibrium. That is, we want the closed-loop system to be
an EL system with EL parameters {T ( q, q ), Vd ( q ), Fd ( q )} .
Since we know that a minimum of the potential energy corresponds to a stable

equilibrium point, the “new” potential energy function should have a global and unique
minimum at the desired position. The following choice is not intuitively obvious but it
is well known in the literature of mechanical systems:
1
Vd (q) = V (q) + k p [q − δ (q∗ )]2 , (18)
2
where by setting the constant
1
δ (q∗ ) = q∗ − g (q∗ )
kp
∗
we ensure that Vd ( q ) has a minimum at q = q . It can be easily shown by
∂ 2 Vd
evaluating (q) , that the function can be made strictly convex and the minimum is
∂q 2
global and unique if moreover k p > mgl . Thus, the energy shaping part of the control
law is given by

Nijmeijer
∂
uES = (V (q) − Vd (q)) = −k p q + g (q∗ ) . (19)
∂q
To make this stable equilibrium attractive we choose the desired Rayleigh dissipation
function Fd ( q ) = 1 kd q , kd > 0 which will induce the correct dissipation properties
2
2
to the system. These choices lead to the control
∂ ∂F
u= (V (q ) − Vd (q )) − c (q )
∂q ∂q

uES uDI
= g (q∗ ) − k p q − kd q. (20)
The closed loop system (17), (20) is a fully-damped Euler-Langrange system with EL
parameters {T ( q, q ), Vd ( q ), Fd ( q )} .
To better understand the passivity property of this controller, let us analyze the passivity
of the closed loop system which we will write as
ml 2 q = −k p q + uDI . (21)
Next, let us consider the energy function of the closed loop system, which is
H(q, q ) = Hp (q, q ) − V (q ) + Vd (q ) . (22)
The total time derivative of H( q, q ) , using (21) yields
H (q, q ) = qu
DI . (23)
That is, the energy shaping control input uES has placed the Lyapunov stable
equilibrium of the pendulum at a desired position q∗ while conserving the passivity
properties of the system. Notice that by integrating the equality above from 0 to T we
can conclude that the closed loop defines a passive map uDI 6 q and, moreover, if the
input uDI ∈ L2 then the system will also be L2 stable.
Next, recalling that the interconnection of passive systems is passive, let us reconsider
the selected input uDI = − kd q . Notice that this a static ISP map q 6 kd q . Hence,
the closed loop system can be regarded as the negative feedback interconnection of the
passive map uDI 6 q with the ISP map q 6 kd q and, as we know, the closed loop is

Nijmeijer
also passive. Strictly speaking, if we add an external input v to (21) we will obtain,
using (21) and (23),
≥ −H (q(0), q (0)) + kd q

2
v | q 2T 2T
.
That is, the map v 6 q is OSP. From here and Lemma 2 we conclude that, if v = 0 (i.e.
the system (21)) then q (t ) → q∗ as t → ∞ .
4.2.2 Lyapunov Stability of the ES+DI Controllers
We have seen that the controllers above render the closed loop system OSP, with output
being the generalized velocities. Notice that this is also the output of the original passive
map defined by the pendulum system alone. To explore the Lyapunov stability of the
closed loop is now a simple task (for these particular examples but in general, it is rather
hard to conclude). One simply has to use the zero state detectability property defined in
Section 3.2 in the following manner:
Set the output q ≡ 0 in the closed loop equation (21). For the first controller, we obtain
that k p q = 0 and hence q = q∗ therefore, the system is zero-state detectable. For the
second controller, we obtain that k p q + g ( q ) − g ( q∗ ) = 0 and be design, q = q∗ is the
only solution to this equation since the origin is the only equilibrium of the system. In
other words, q ≡ 0 implies that q = q∗ and therefore, the system is zero-state detectable.
Notice next that for either case, we can perform a loop transformation so that the closed
loop system can be regarded as two OSP maps. This can be accomplished by simply
“redefining” the control inputs uDI = −0.5kd q and u ES = g ( q ) − k p q − 0.5 k d q for
the first controller and, uES = g ( q∗ ) − k p q − 0.5kd q for the second one. Notice that
this leaves the system unchanged. Finally, we can invoke item (iii) of Proposition 2.
4.3. EL Controllers
Two important characteristics of the PD controller of the previous section are that these
controllers are firstly, that they are static and secondly, they preserve the passivity
structure of the system. However, they present as well several drawbacks which stymie
their utilization in some applications:
1. Measurement of the generalized velocities q , and full actuation are required to

add the necessary damping.
2. No amplitude constraints are imposed on the control input.
3. The potential energy function V (q ) is supposed to be exactly known.

Nijmeijer
We will not treat in much detail all these drawbacks since that are dealt with in some of
the references on EL systems included in the Bibliography using a technique we will
illustrate here by addressing only the first issue. We will present a class of structure
preserving passive controllers. To that end, we must remark the fundamental fact that
the feedback interconnection of two EL systems yields also an EL system. More
precisely we have the following
Proposition 4 (Interconnected EL systems): Consider two EL systems

Σ p : {T p ( q p , q p ), Vp ( q p ), F p ( q p ), M p } and Σ c {T c ( qc , qc ), Vc ( qc , q p ), Fc ( qc )}
with generalized coordinates q p ∈ \ and qc ∈ \ , respectively, (notice that the
np nc
potential energy of Σ c depends on q p ). Interconnect the systems via
∂Vc (qc , q p )
Mp u = − ,
∂q p
where u is the input of the subsystem Σ p . Under these conditions, the closed-loop
system is an EL system Σ :{T ( q , q ), V ( q ), F ( q )} , with generalized coordinates
q := [q p , qc ] and EL parameters
T (q, q ) = T c (qc , qc ) + T p (q p , q p ) V (q ) = Vc (qc , q p ) + Vp (q p ) ,
F = Fc (qc ) + Fp (q p ) .
Based on this proposition we can now proceed to construct a position feedback

controller of EL systems. The technique is simple. Let Σ p be a given EL system for
which we know the EL parameters, in particular, we know the expression of the total
energy of the system. Assume that only the generalized positions q p are available from
measurement, the system is fully-actuated i.e., Mp = I and, undamped i.e., Fp ≡ 0 .

Nijmeijer
Figure 3: Interconnection of the EL system and the EL controller.
We will realize the control design in the following steps:
1. Taking into account that Σ p : u 6 q p is passive, we will close a first loop as shown
in Figure 3. It is not difficult to see that the input to Σ p in this case, corresponds to the
control uES = − k p [q p + δ ( q∗ )] . Notice also that Σ′p is a new EL system with EL
parameters
1
{T p (q p , q p ), [q p + δ(q∗ )] K p [q p + δ(q∗ )],0, I } .
2
2. Secondly, we introduce an adequate damping through an EL controller, that is, let us
consider a dynamical position feedback controller Σ c with EL parameters
⎧ 1 2 1 ⎫
⎨0, Vc (qc , q p ) := qc + bq p , qc2 ,0 ⎬ .
⎩ 2b ab ⎭
The internal dynamics of this controller is derived using the EL equations to obtain
qc = − a (qc + bq p ) .
Some readers will notice that this EL system has zero kinetic energy. While this is
physically impossible, we call this controller EL since as we show next, the closed loop
system is an EL system moreover, the kinetic energy of the controller does not play any

Nijmeijer
role in the stabilization task here since the control goal is to achieve a constant set-point.
3. According to Proposition 4 the system depicted in Figure 3 is also Euler-Lagrange

and, more over, its EL parameters are
⎧ 1 1 2 ⎫
⎨T ( q, q ) = T p (q p , q p ), V ( q ) = [ q p + δ ( q∗ )] K p [ q p + δ ( q∗ )] + Vp ( q p ), F ( q ) := qc ,0 ⎬
⎩ 2 ab ⎭
and the dynamics of the closed loop system can be derived using the EL equations.
To analyze the Lyapunov and input-output stability of the closed loop system we follow
a passivity approach. Once more, we will exhibit the fact that the inputs and outputs
defining the passive maps do not necessarily correspond to the physical measurable
ones! Let us consider now the system of Figure 4. Notice that it is equivalent to the
system of Figure 3. We can draw the following conclusions.
Figure 4: Passive EL system and EL controller interconnection.
1. The system Σ1 : v 6 q p is passive because of the following:

Σ p : v − uES 6 q p is passive and the integrator block in the feedback loop is
also passive. To see the latter, integrate
q p (t )uES (t ) from 0 to T , to obtain that
q p | uES ≥ β := − K p q p (0) 2 − [ q p (T ) − q p (0)]δ ( q∗ ) .
2. The filter in the outer feedback loop, defines an OSP map Σ 2 : q p 6 uDI since
the filter is strictly positive real. The latter can be seen by substituting the
derivative operator p , by jw and observing that ReH ( jw ) > 0 for all
∞ > w > 0 . Alternatively, we can employ the storage function

Nijmeijer
a 2
Hc (uDI ) := uDI
2b
and evaluate its total time derivative using the dynamics of the filter in the block Σ 2 ,
i.e., uDI = auDI − bq p to obtain H (u ) := −au − q u . The OSP property
2
c DI DI p DI
follows by integrating on both sides of the latter equation as done before.
We have thus constructed a controller which has the same physical properties than the
plant: an EL controller. One of the obvious advantages of this type of controller is that it
preserves the structure of the system while using only the available measurement. Form
a passivity point of view, the example above shows that one can inject the appropriate
damping through a dynamic system. The filter chosen here is also known as dirty
derivatives and it is widely used in applications.
4.4. Tracking a Time-varying Reference
The passivity property described above is sufficient to solve regulation tasks in

mechanical systems, where the PBC only needs to modify the potential energy and the
dissipation function. However, to study tracking problems or deal with other EL
systems of electrical or electromechanical nature, we need a stronger property. The
main reason is that in these cases a desired behavior should be imposed, not only on
q as well, which in its turn translates into the need for modifying the kinetic energy.
We shall not explore in detail this passivity-based approach but only describe an
algorithm which allows us to track a time-varying reference. The case we present here
also shall illustrate the fact that, since passivity is an input-output property, it can be
used also for time-varying nonlinear systems.
The passivity-based approach we describe next consists on designing (seeking for) a

controller which makes the closed loop system take the form
D ( q ) s + [C ( q, q ) + K d ( q, q )]s = 0 , (24)
where s denotes a virtual error signal that we want to drive to zero,

K d ( q, q ) = K d (q, q ) > 0 is a damping injection matrix, and C (q, q ) is a matrix,
univocally defined by the inertia matrix D ( q ) and which satisfies
D ( q ) = C ( q, q ) + C ( q, q ) . (25)
Notice that (25) is equivalent to the skew-symmetry property
z [ D (q ) − 2C (q, q )]z = 0, ∀ z ∈ \ n
which holds for many EL systems. As a matter of fact, under a suitable factorization one

Nijmeijer
can express the EL equations (8) in the form
D(q )q + C (q, q )q + g (q) = 0 , (26)
where g ( q ) := ∂∂Vq .
The motivation for aiming at (24) stems from the following important fact: the
differential equation
D ( q ) s + [C ( q, q ) + K d ( q, q )]s = Ψ ,
where D ( q ) and K d ( q, q ) are positive definite and C (q, q ) satisfies (25) defines an
output strictly passive operator Σ d : Ψ 6 s . Consequently, if Ψ ∈ L2 we have
s ∈ L2 .
To see this, one can use the storage function
1
Hd = s D(q) s ≥ 0 (27)
2
and differentiate it with respect to time then, using the skew-symmetry of
D (q ) − 2C (q, q ) we obtain that Hd ≤ − K d ( q, q ) s 2 + Ψ s . The OSP property
follows by integrating on both sides of this inequality from 0 to T . The second part of
the proof follows from the fact that OSP systems are L2 -stable. Henceforth, with
Ψ ∈ L2 we have s ∈ L2 .
An important consequence of this is that, since the system Ψ 6 s is made OSP then,
when Ψ ≡ 0 we have that the output s → 0 as t → ∞ . Therefore, if one is interested
in the convergence of the variables q and q to time-varying reference trajectories
q∗ q∗ one has simply to satisfy the zero-state detectability condition. A simple choice is
then s := q − q∗ + λ ( q − q∗ ) since this implies that q = −λ ( q − q∗ ) + s .
Finally, we can calculate the control input u which makes the EL system take the form
(24). This is given simple by the difference between the latter and (26) i.e.,
u = g(q) − Kd [q − q∗ + λ(q − q∗ )] + C(q, q)[q∗ − λ(q − q∗ )] + D(q)[q∗ − λ(q − q∗)] + Ψ .
5. Epilogue
We have given a short introduction in one of the more systematic ways of controlling
systems that possess an energy dissipation property. More specifically, this amounts to
saying that at least as much energy is fed into the system as is stored in the system. In

Nijmeijer
the context of passivity the notion of energy can be far more general than the physical
energy concept as for instance encountered in RLC circuits or simple mechanical
systems. The passivity of a system is often a desirable property and may for instance
induce stability of an equilibrium. Because of this, one of the systematic controller
design methods is based upon the idea of making the closed-loop system passive. The
passivity condition is of great importance in practical cases and may help in stabilization
and tracking of systems.
Glossary
Equilibrium: It is a constant value of the same dimension as the state and such
that, if the latter equals to the equilibrium at an instant t, the
state will not change for any later time.
Generalized It is a coordinate which may have different physical units. If it is
coordinate: a vector, then each element may have different units.
Input: An input to a system is an external stimulus which acts upon a
system modifying it in some manner.
Lyapunov stability: A the equilibrium of a system is said to be Lyapunov stable if
the state values remain arbitrarily close to the equilibrium at any
instant of time, provided that the initial states were “sufficiently
close” to the equilibrium. The equilibrium is said to be
asymptotically Lyapunov stable if, moreover the state values
converge to the equilibrium as we consider the time to go to
infinity.
Origin: Is the zero value of the state.
Output: The output of a system is the result of the “changes”
experienced by a system stimulated by an input.
State: The state is a set of variables which fully describe the
configuration of a system at a given time instant. These
variables may have a physical meaning or not.
Bibliography
V. Arnold. (1989) Mathematical Methods of Classical Mechanics. New York :Springer-Verlag .

H. Berghuis and H. Nijmeijer. (1993) A passivity approach to controller-observer design for robots. IEEE
Trans. On Robotics Automat., 9-6:740-755.
C.Byrnes, A. Isidori, and J.C. Willems. (1991) Passivity, feedback equivalence, and the global
stabilization of minimum phase nonlinear systems. IEEE Trans. on Automat. Contr., AC-36(11):1228-
1240.[Section 3 is based on this work]
C.A. Desoer and M. Vidyasagar. (1975) Feedback Systems: Input-Output Properties. 266p. New York
:Academic Press, .[Input-Output stability theory is thoroughly studied. The material of Section 2 is based
on this work]
H. Goldstein. (1974) Classical Mechanics. 672p. Addison Wesley.[An excellent text on mechanics]
D. Hill and P. Moylan. (1976)The stability of nonlinear dissipative systems. IEEE Trans. on Automat.
Contr., Pages 708-711.[Fundamental results on interconnection of passive systems]
D. E. Koditschek. Natural motion of robot arms. In Proc. 23rd. IEEE Conf. Decision Contr., Las Vegas,
NV., 1984.[Passivity based control of EL systems and in particular mechanical systems.]
A Loría R. Kelly, R. Ortega, and B. Santibñez. (1996) On output feedback control of Euler-Lagrange

Nijmeijer
systems under input constraints. IEEE Trans. Automat. Contr. , 42(8):1138-1142.[ Passivity based control
of EL systems and in particular mechanical systems]
A. Loría and H. Nijmeijer. (1998) Nonlinear control systems: (Output feedback) design methods. In John
G. Webster, editor, Wiley encyclopedia of Electrical and Electronic Engineering. John Wiley & Sons IC.
. Article #1025 in Section “Automatic Control”[ Passivity based control of EL systems and in particular
mechanical systems]
A. Loría and H. Nijmeijer (1998) : Output feedback tracking control of Euler-Lagrange systems via
bounded controls. Syst. & Contr. Letters, 33(3):151-163.[ Passivity based control of EL systems and in
particular mechanical systems]
R. Lozano, B. Brogliato, O. Egeland, and B. Maschke.(2000) Dissipative systems analysis and control.
Communications and control Engineering. London : Springer Verlag. [Passivity based control of
nonlinear systems].
A. M. Lyapunov. Problème de la stabilitè de mouvement. (1893) Annales de la facultè de sciences de
Toulouse, 9:203-474, 1907. (Translation from the original published in Comm. Soc. Math. Kharkov 1893,
reprinted in Ann. Math. Studies 17, Princeton 1949). See also “Stability of motion”, Academic Press: NY
1996..[Liapunov stability was launched]
I. M. Y. Mareels and D. Hill. Monotone stability of nonlinear feedback systems. J. Math. Systems’
Estimation and Control, 2:275-291, 1992.[ Fundamental results on interconnection of passive systems]
R. Ortega, A. Loría R. Kelly, and L. Praly.(1995) On passivity-based output feedback global stabilization
of Euler-Lagrange systems. Int. J. Robust and Nonlinear Control, Special issue on Control of nonlinear
mechanical systems, 5(4):313-325. (H. Nijmeijer and A. J. van der Schaft eds.).[ Passivity based control
of EL systems and in particular mechanical systems]
R. Ortega, A. Loría P.J. Nicklasson, and H. Sira-Ramírez. (1998) Passivity-based Control of Euler-
Lagrange Systems: Mechanical, Electrical and Electromechanical Applications. Communications and
control Engineering. London : Springer Verlag . ISBN 1-85233-016-3.[ Section 4 is based fully on this
work. ]
R. Ortega and M. Spong. (1989) Adaptive motion control of rigid robots: A tutorial. Automatica, 25-
6:877-888.[This tutorial introduces the term PBC]
R. Sepulchre, M. Janković, and P. Kokotović. (1997) Constructive nonlinear control. Springer
Verlag.[Passivity based control of nonlinear systems]
J. J. Slotine and W. Li. (1988) Adaptive manipulator control: a case study. IEEE Trans. on Automat.
Contr., AC-33:995-1003.[ Passivity based control of EL systems and in particular mechanical systems]
M. Takegaki and S. Arimoto. (1981) A new feedback method for dynamic control of manipulators. ASME
J. Dyn. Syst. Meas. Contr., 103:119-125.[The so-called energy-shaping plus damping injection technique
was introduced in the context of robot control]
A. J. van der Schaft.(1999) L2 -Gain and passivity techniques in nonlinear control. Communication and
control Engineering. Heidelberg : Springer Verlag, 2nd edition.[The material of Sections 2 and 3 is based
on this work. Theorems 1 and 2 are from this work.]
J. C. Willems. Dissipative dynamical systems. Part I (1972): General theory. Arch. Rat. Mech . and
Analysis, 45(5):321-351.[The theory of dissipative systems is introduced]
J. C. Willems. Dissipative dynamical systems. Part II (1972): Linear systems with quadratic supply rates.
Arch. Rat. Mech. and Analysis, 45(5):352-393.[ The theory of dissipative systems is introduced]
Biographical Sketches
Antonio Loria was born in Mexico in 1969. He got the BSc degree in Electronic Engineering from the
ITESM, Monterrey, Mexico in 1991. He got the MSc and PhD degrees in Control Engg. from the UTC,
France in 1993 and Nov. 1996 respectively. From December 1996 thru Dec. 1998, he was successively an
associate researcher at Univ. of Twente, The Netherlands; NTNU, Norway and the CCEC of the Univ. of
California at Sta Barbara, USA. Dr. Loria is currently ``Charge de Recherche'', at the the French National

Nijmeijer
Centre of Scientific Research (CNRS). He is (co)author of more than 75 scientific articles and the book
``Passivity based control of Euler-Lagrange systems'' Springer Verlag, 1998. Antonio Loria is associate
editor of Systems and Control Letters. His research interests include: modeling and control of Euler-
Lagrange systems, stability analysis of nonlinear time-varying systems, biped locomotion, and output
feedback stabilization. Detailed information and publications are available at:
http://public.lss.supelec.fr/perso/loria.
Henk Nijmeijer (1955) obtained his MSc-degree and PhD-degree in Mathematics from the University of
Groningen, Groningen, the Netherlands, in 1979 and 1983, respectively. From 1983 until 2000 he was
affiliated with the Department of Applied Mathematics of the University of Twente, Enschede, the
Netherlands. Since, 1997 he was also part-time affiliated with the Department of Mechanical Engineering
of the Eindhoven University of Technology, Eindhoven, the Netherlands. Since 2000, he is full-time
working in Eindhoven, and chairs the Dynamics and Control section. He has published a large number of
journal and conference papers, and several books, including the 'classical' Nonlinear Dynamical Control
Systems (Springer Verlag, 1990, co-author A.J.van der Schaft). Henk Nijmeijer is editor in chief of the
Journal of Applied Mathematics, corresponding editor of the SIAM Journal on Control and Optimization,
and board member of the International Journal of Control, Automatica, European Journal of Control,
Journal of Dynamical Control Systems, SACTA, International Journal of Robust and Nonlinear Control,
and the Journal of Applied Mathematics and Computer Science. He is a fellow of the IEEE and was
awarded in 1987 the IEE Heaviside premium.

Passivity Based Control

Uploaded by

Copyright:

Available Formats

Passivity Based Control

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Passivity Based Control

Uploaded by

Copyright:

Available Formats

Eolss Publishers Co. Ltd.

Copyright © 2009 EOLSS Publishers/ UNESCO

Rev 1.1, October 2009

Information on this title: www.eolss.net/eBooks

ISBN 978-1-84826-152-5 (e-Book Adobe Reader)

British Library Cataloguing-in-Publication Data

Library of Congress Cataloging-in-Publication Data

PASSIVITY BASED CONTROL

Keywords: Passivity, automatic control, mechanical systems, robot manipulators,

Passivity is a fundamental property of many physical systems which may be roughly

©Encyclopedia of Life Support Systems (EOLSS) 206

A fundamental property of passive systems is that, regarding a feedback interconnection

©Encyclopedia of Life Support Systems (EOLSS) 207

2. Passivity: Mathematically Speaking

2.1. In a General Input-Output Framework

Definition 1 ( L2 and L2 norms): The L2 norm of a signal f : \ ≥ 0 → \ , f (t ) is

denoted f (t ) 2T and defined by

©Encyclopedia of Life Support Systems (EOLSS) 208

Definition 2 ( L2e -space): We say that f : \ ≥ 0 → \ belongs to L2e if and only if

For illustration we may consider u to be an input voltage supplied to an electrical load

With these tools we can now properly define the following

Definition 4 (Passivity): An operator H : u 6 y is passive if there exists a β ∈ \ such

Definition 5 (Output Strict Passivity): An operator H : u 6 y is output strictly

©Encyclopedia of Life Support Systems (EOLSS) 209

Definition 6 (Input Strict Passivity): An operator H : u 6 y is input strictly passive

Passive interconnected systems

Theorem 1: Consider the input-output system depicted in Figure 1. Let

Figure 1: Feedback interconnection of passive systems

©Encyclopedia of Life Support Systems (EOLSS) 210

• If ∑1 : e1 6 y1 is passive and ∑ 2 : y1 6 y2 is ISP or

©Encyclopedia of Life Support Systems (EOLSS) 211

Figure 2: Loop transformed feedback interconnected system

hence, we must impose k < δ i 2 .

3. Stability of Passive Systems

©Encyclopedia of Life Support Systems (EOLSS) 212

A “relaxed” definition of L2 stability is that an operator is L2 stable if it maps

Definition 7 ( L2 -stability): The state space system ∑ is said to be L2 stable with

Proposition 1: If ∑ : u 6 y is OSP then it has finite L2 -gain.

3.2. From L2 -Stability to Lyapunov Stability

Definition 8 (Zero-state detectability): A state-space system

Definition 9 (Zero-state observability): A state-space system

Roughly speaking, thinking of physical systems these properties can be regarded as

©Encyclopedia of Life Support Systems (EOLSS) 213

Lemma 1: Let y = G ( p )u , where G ( p ) is an n × m strictly proper, stable transfer

Lemma 2: If the system ∑ : u 6 y is OSP, y ∈ L2 ∩ L∞ , y (t ) ∈ L∞ and, the system

are passive or output strictly passive, with storage functions H1 ( x1 ) and

then the following holds true.

©Encyclopedia of Life Support Systems (EOLSS) 214

4. PBC of Euler-Lagrange Systems

Without further introduction let us consider a dynamical system with n degrees of

system is an Euler-Lagrange system if its dynamics is described by the equations

©Encyclopedia of Life Support Systems (EOLSS) 215

where D ( q ) ∈ \ n × n is the generalized inertia matrix that satisfies

In summary the external forces can be expressed by

where Qς is an external signal that models the effect of disturbances.

As explained in the Introduction, in PBC the control objective is achieved by imposing

Definition 10 (EL equations and EL parameters): The EL equations of motion

©Encyclopedia of Life Support Systems (EOLSS) 216

{T ( q, q ), V (q ), F ( q )} in the case when only dissipative forces affect the EL system.

Definition 11 (Underactuated EL systems.): An EL system is fully-actuated if it has

Lemma 2: If the system ∑ : u 6 y is OSP, y ∈ L2 ∩ L∞ , y (t ) ∈ L∞ and, the system

{T ( q, q ), V (q ), F ( q )} in the case when only dissipative forces affect the EL system.

u | M q ≥ H[q(T ), q (T )] − H[q(0), q (0)] (14)

3. We see that damping can be easily added-along the actuated channels- if q is

= g (q∗ ) − k p q − kd q. (20)

ml 2 q = −k p q + uDI . (21)

H(q, q ) = Hp (q, q ) − V (q ) + Vd (q ) . (22)

The total time derivative of H( q, q ) , using (21) yields

≥ −H (q(0), q (0)) + kd q

1. Measurement of the generalized velocities q , and full actuation are required to

T (q, q ) = T c (qc , qc ) + T p (q p , q p ) V (q ) = Vc (qc , q p ) + Vp (q p ) ,

1. The system Σ1 : v 6 q p is passive because of the following:

q p (t )uES (t ) from 0 to T , to obtain that

q p | uES ≥ β := − K p q p (0) 2 − [ q p (T ) − q p (0)]δ ( q∗ ) .

D(q )q + C (q, q )q + g (q) = 0 , (26)