control and computation -...
TRANSCRIPT
Members: Mato Baotic
Francesco Borrelli
Tobias Geyer
Pascal Grieder
Frank Christophersen
Fabio Torrisi
Affiliates: Alberto Bemporad (Univ. of Siena)
Giancarlo Ferrari-Trecate (INRIA)
Control and ComputationControl and Computation
ETH ActivitiesETH Activities
Manfred Morari
ETH Zürich
Control EngineeringGroup
Walter SchaufelbergerAdolf H. Glattfelder
Franta J. Kraus
System Dynamics & Control Group
Manfred Morari
AdministrationInfrastructure
Teaching Program
FachgruppenFachgruppen
ETH Zürich
People (November 2000)People (November 2000)
FTE
Professors 2 (+1) 2
Researchers 11 10
Graduate Students 14 14
Visitors 3 3
Administrative/Technical Staff 7 4.6
TOTAL 37 33.6
ETH Zürich
System Dynamics & ControlSystem Dynamics & ControlBiomedical Systems
- Rehabilitation Engineering Popovic, Jezernik, Keller, Pappas
- Anesthesia Glattfelder, Gentilini, Stadler
Active Noise and Vibration Control, Kaiser, Seba
Intelligent Materials
Nonlinear Dynamics of Distillation Columns Bonanomi, Dorn, Ulrich
Hybrid Systems Bemporad, Borrelli, Cuzzola, Ferrari, Kojima,
- Analysis: Controllability, Verification, Stability, ... Mignone, Torrisi
- Synthesis: Control, Estimation, Fault Detection
- Applications
Model Development for Control Pearson
- Nonlinear Model Reduction
- Nonlinear System Identification
Role of ETH in CCRole of ETH in CC
• Controller Synthesis via Optimization
• ABB Case Study
• Coordination of Case Studies
Truth value operator:
From Algebraic Equalities to From Algebraic Equalities to MixedMixed--Integer Linear InequalitiesInteger Linear Inequalities
Xb c 2 f0;1gP(X1; . . .; Xn)b c = 1 Aîöô B
îö= î1; . . .; în[ ]0 2 f0; 1gn
z = îxî 2 f0; 1gx 2 [m;M]
ú z ô Mîz õ mîz ô x àm(1 à î)z õ x àM(1 à î)
8><>:
x ô 0b c = î x ôM(1 à î)x õ ï+ (m + ï)î
ú
Mixed product
Propositionallogic
Threshold condition
Algebraic equalities MI linear inequalities
(Williams, 1977)
(Glover, 1975)(Witsenhausen, 1966)
MLD Hybrid ModelsMLD Hybrid Models
x(t + 1) = Ax(t) +B1u(t)y(t) = Cx(t) +D1u(t)
Mixed Logical Dynamical (MLD) form
+B2î(t) +B3z(t)+D2î(t) +D3z(t)
E2î(t) + E3z(t) ô E4x(t) + E1u(t) + E5
(Bemporad, Morari, Automatica, March 1999)
• Well-Posedness Assumption :
are single valuedWell posedness allows defining trajectories in x- and y-space
î(t) = F(x(t); u(t))z(t) = G(x(t); u(t))
fx(t); u(t)g ! fx(t + 1)gfx(t); u(t)g ! fy(t)g
x; y; u = ?c?`
ô õ; ?c 2 Rnc; ?` 2 f0; 1gn`; z 2 Rrc; î 2 f0; 1gr`
Major Advantage of PWA/MLD Framework
All problems of analysis:• Stability• Verification• Controllability / Reachability• Observability
All problems of synthesis:• Controller Design• Filter Design / Fault Detection & Estimation
can be expressed as (mixed integer) mathematical programmingproblems for which many algorithms and software tools exist.However, all these problems are NP-hard.
Finite Time Optimal ControlFinite Time Optimal Control
Predicted outputs
Manipulated (t+k)u
Inputs
t t+1 t+m t+p
futurepast
Jopt =minUJ(U,x(t)) :=P
k=0
N kQ(x(t+ k|t)àxö)k+ kRu(t+ k)k
U := {u(t), u(t+1), . . ., u(t+Nu)}
subject to umin ô u(t+k) ô umax
xmin ô x(t+k|t) ô xmax
system dynamics
Model Perf. index Method
Linear Quadratic ProgramLinear Linear ProgramHybrid Mixed Integer Quadr. ProgramHybrid Mixed Integer Linear Program
Finite Time Optimal ControlFinite Time Optimal Control
U := {u(t), u(t+1), . . ., u(t+Nu)}
subject to umin ô u(t+k) ô umax
xmin ô x(t+k|t) ô xmax
Jopt =minUJ(U,x(t)) :=P
k=0
N kQ(x(t+ k|t)àxö)k+ kRu(t+ k)k
system dynamics
k á k∞k á k2k á k∞
k á k2
Feedback Control SolutionFeedback Control Solution
U := {u(t), u(t+1), . . ., u(t+Nu)}
subject to umin ô u(t+k) ô umax
xmin ô x(t+k|t) ô xmax
Jopt =minUJ(U,x(t)) :=P
k=0
N kQ(x(t+ k|t)àxö)k+ kRu(t+ k)k
system dynamics
Theorem: The solution of the finite time optimalcontrol problem yields a piecewise affinestate feedback control law
Computation of Feedback Control LawComputation of Feedback Control Law
U := {u(t), u(t+1), . . ., u(t+Nu)}
subject to umin ô u(t+k) ô umax
xmin ô x(t+k|t) ô xmax
Jopt =minUJ(U,x(t)) :=P
k=0
N kQ(x(t+ k|t)àxö)k+ kRu(t+ k)k
system dynamics
• The feedback controller can be computed explicitly
• Off-line optimization: optimize for all x(t)
Multi-parametric Program
Solvers for LP, QP, MILP and MIQP? are available
• Traction control (Ford Research Center )
• Gas supply system (Kawasaki Steel )
• Batch evaporator system (Esprit Project 26270 )
• Anesthesia (Hospital Bern )
• Hydroelectric power plant ( )
• Power generation scheduling ( )
ApplicationsApplications
Current and Future ProjectsCurrent and Future Projects
• Algorithm Improvements (mpMIQP)
• Controller Characterization and Storage
• Controller Simplification- Order reduction
- Decentralized structure
• Robustness