Modeling, control and observation of
dynamical systems case of Systems
of Systems (SoS)
Ali Charara
Reine Kfoury Talj
[email protected][email protected] Program
Unified representation, examples of modeling
Analysis of systems properties : stability, controllability, observability,
Linearization
Linear systems : control and observer
State feedback
Luenberger observers
Input-output decoupling control
Nonlinear control
Global linearization (input-output decoupling)
Lyapunov
Introduction to Systems of Systems control
Hierarchical control
Decentralized and Distributed control
Networked control
Examples and case study (Matlab)
Systme de Systmes
Examples of Systems of Systems
Multimodal transport services
Systme de Systmes
Examples of Systems of Systems
E-Health
Systme de Systmes
Examples of Systems of Systems
Natural disasters management
Examples of Systems of Systems
Examples of Systems of Systems
What are Systems of Systems?
Existing Complex Systems
Exclusive, Autonomous, Local
Transformation
(Keating, et al., 2003)
What are Systems of Systems?
System of Systems
Integrated,
Aligned,
and Mission,
Interconnected,
Integrated
Transforming
Global, Emergent Structure
(Keating, et al., 2003)
A schematic representation of a system of systems
(Samad, Parisini, Part3-IEEE CSS)
Some Definitions for Systems of Systems (SoS)
Systems of Systems are large-scale integrated systems which are
heterogeneous and independently operable on their own, but
are networked together for a common goal. The goal may be
cost, performance, robustness, etc
A System of Systems is a "super system" comprised of other
elements which themselves are independent complex
operational systems and interact among themselves to achieve
a common goal. Each element of a SoS achieves wellsubstantiated goals even if they are detached from the rest of
the SoS.
(Karcanias, 2011)
Characteristics of Systems of Systems
Operational independence of component systems
Managerial independence of component systems
Geographical distribution
Emergent behavior
Evolutionary development processes
(Maier, 1998)
Some application domains of SoS
Air Traffic Control
Internet
Intelligent Transport Systems
Renewable energy systems
Robotic swarms
Space
Defense and military
Environment, etc
Large Scale Systems vs Systems of Systems
TOP
LSS
TOP
BOT.
BOT.
LSS
Traditional LSS Modeling
SoSE Modeling Difficulty
An introduction to Systems of Systems Control
Modeling: Hard to find a simple mathematical formalism to define all systems
aspects.
Control: Every case have its own study and law.
Communication: Systems operate using different languages and semantics.
Hierarchical Control
Decentralized Control
Distributed Control
Consensus-based Control
Networked Control
An introduction to Systems of Systems Control
Hierarchical control
An introduction to Systems of Systems Control
Hierarchical control
A system is chosen to play the role of a coordinator
Each system receives its N-1 neighbors data via wireless communication
An optimal controller is designed for each system
The coordination of the N solutions via an iterative algorithm gives an
optimal solution of the SoS
An introduction to Systems of Systems Control
Centralized control
Decentralized control
Taken from the theory of large-scale (complex) systems, one can
share the control action among a finite number of local controllers.
An introduction to Systems of Systems Control
Decentralized control
The sensory information of SoS is distributed across the N domains
N local controllers are designed to meet systems criteria
A global control component reacting to neighboring (N-1) systems is
added to the local control strategies.
A cost function can be chosen for the local design problems
An introduction to Systems of Systems Control
Distributed control
An introduction to Systems of Systems Control
Consensus-based Control
It is a cooperative control paradigm based on consensus among
systems in a SoS.
Multiple rovers
No single control unit/rover
Units must agree on goal
Sub goals may be different for each unit
Shared communications
(Ren and Baird 2008)
An introduction to Systems of Systems Control
Consensus-based Control
Communications
How well can the robots talk to each other?
The communication among systems have limited bandwidth,
connectivity and questions arise like what and when and to
whom a system should communicate?
Constrained: 1 cannot talk to 3 directly
Ideal: All 3 talk directly
An introduction to Systems of Systems Control
Consensus-based Control
Step1 : objective
An introduction to Systems of Systems Control
Consensus-based Control
Step2 : Coordination Variables
(Mo Jamshidi)
An introduction to Systems of Systems Control
Consensus-based Control
Step3 : Centralized strategy
An introduction to Systems of Systems Control
Consensus-based Control
Step4 : Consensus Building
An introduction to Systems of Systems Control
Networked control
An introduction to Systems of Systems Control
Networked control
One of the main challenges in NCSs is the loss or
delays in transmission and receipt of data from sensors to
controllers and from controllers to actuators.
The challenge in SoS networked control is to develop an SoS
distributed control system which can tolerate lost packets,
partially decoded packets, delays, and fairness issues i.e.
add robustness to the control paradigm.
An introduction to Systems of Systems Control
Hierarchical control
A system is chosen to play the role of a coordinator
Each system receives its N-1 neighbors data via wireless communication
An optimal controller is designed for each system
The coordination of the N solutions via an iterative algorithm gives an
optimal solution of the SoS
Hierarchical control
1. Model-Coordination Method
Consider the following optimization problem:
Minimize
, ,
, ,
Subject to
= 0.
where is the state vector, control variables, is a vector of interaction between
subsystems, an objective function, and () a constraint function.
Let the problem and its objective function be decomposed into two subsystems:
and
, ,
,
,
+
= 0,
( ,
= 1,2
Hierarchical control
1. Model-Coordination Method
First-level problem subsystem
min
Find
=
( , , )
,
subject to ( , , , ) = 0
Second-level problem
min
=
+
( )
Hierarchical control
2. Goal-Coordination Method
Consider the same optimization problem: Minimize
, ,
Subject to
, ,
= 0.
In this method, the interaction is completely removed by cutting all the links among the
subsystems . Let be the outgoing variable from the th subsystem, and the
corresponding input. Since the interaction is removed, its obvious that .
The global problem is completely decoupled; the subsystems problems are completely
independent. The new formulation of the objective functions is:
, , ,
=0
, , ,
=0
It is necessary that the interaction balance principle be satisfied ( = ).
Consider the following new cost function:
, , , , =
, ,
+
, ,
+ ( )
Where is a vector of weighting parameters which causes any interaction
unbalance to affect the objective function.
Hierarchical control
2. Goal-Coordination Method
First level problem:
Subsystem 1:
, ,
min
,! ," ,#$
Second level problem (coordinator):
Subsystem 2:
min
,
subject to
$ ,!$ ,"$ ,#
subject to
=0
=0
min & = min( )
%
Hierarchical control
To apply these methods, two principles can be applied: the Interaction prediction
principle and the interaction balance principle.
Consider the following large-scale linear time-invariant system:
' =( ) +* ) ,
)+ = +
Where ,-. are - 1 ,-. 0 1 state and control vectors. It is assumed that (the
above system) can be decomposed into
' = (
) +*
Where the interaction vector
) +1
) ,
0 =
= 45 6
68
69
+,
= 1,2, , 3
is a linear combination of the states of the other 3 1 subsystems, and 5 6 is an - -6
matrix.
Hierarchical control
Example
Consider the following 12th-order system
0
1
0
0
0
1
3 2 1
0
0
0
0
0
0
1
0
0
'=
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
3 2 0
1
0
0
0
0
1
0
0
0
0
0
1
1
0 1 2 3
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
1
2 1
0
0
1
0
0
0
+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
Hierarchical control
This system can be decomposed into 4 3rd-order subsystems with state equations
0
1
0
' = 0
0
1
3 2 1
0
<' = 0
1
1
0
0
1
2 3
0 0
+ 0 0
1 0
0 0
<+ 0 0
0 1
0
1
0
; ' = 0
0
1
1 3 2
0
1
0
< ; =' = 0
3 2
0
1
1
0 0
+ 0 0
0 0
0 0
=+ 0 0
0 0
with the interaction matrices 5 6 , = 1,2,3,4, ? = 1,2,3,4, ? given by:
5
5
5=
=5
=
0
= 0
1
0
= 0
0
<
=5
= 5< = 5<= = 5= = 0;
0 0
0 0 0
0 0 ; 5 = 0 0 0 ; 5
0 0
1 0 0
0 0
0 0 0
0 0 ; 5=< = 0 0 0 .
1 0
1 0 0
<
0 0
= 0 0
0 1
0
0 0 0
0 ; 5< = 0 0 0
0
0 1 0
Hierarchical Control
Hierarchical Control
Interaction prediction principle in Model coordination
Algorithm 1: Interaction prediction method for Continuous-Time Systems.
Step 1: Solve N independent differential matrix Riccati equations:
' ) =
Store
) ( (
) +
) @
) A
)B
)C = A
) ; = 1,2, . . , 3 ,-. 0 ) )C .
Step 2: Initialize an arbitrary value for ()) and find the corresponding value for ()),
then solve N adjoint differential equations:
F' ) = ( @
F )
) 1
Store F ) ; = 1,2, . . , 3 ,-. 0 ) )C .
) + 4 56
68
69
)B F ()C ) = 0
Hierarchical Control
Step 3: Solve ' ) = ( @
)
) @F ) +1
Store ) ; = 1,2, . . , 3 ,-. 0 ) )C .
())
)B
0 =
Step 4: At the second-level (coordinator), use the results of Steps 2 and 3 to update the
coordinator vector:
())
())
GH
1 I ())
= 4 5 6 6 ())
68
69
where J is the number of the current iteration.
Step 5: Check for the convergence at the second-level (coordinator) by evaluating the
overall interaction error:
7
LM
& ) = 4 K
8
) 4 5 6 6 ())
68
69
) 4 5 6 6 ()) .)O)
68
69
if a desired convergence is achieved, stop. Otherwise set J = J + 1 and go back to Step 2.
Hierarchical Control
Hierarchical Control
Hierarchical Control
Optimal states trajectories
Hierarchical Control
Optimal states trajectories
Hierarchical Control
