Professional Documents
Culture Documents
Outline
1 Introduction
2 Proposed Strategy
3 Analysis
4 Illustrative Example
5 Conclusion
1 Introduction
2 Proposed Strategy
3 Analysis
4 Illustrative Example
5 Conclusion
Distributed Control
Infrastructure systems
• examples: water distribution networks, electric
power system, building automation
• spatially distributed
• often large-scale
Centralized control
• simple and well-known approach
• but often infeasible due lack of
communicational/computational power
• undesired due to privacy reasons
Distributed control
• decomposition of the plant in subsystems
• local controller for each subsystem
• communication to improve the performance
x i,t ∈ Xi , u i,t ∈ Ui , ∀t ∈ N
1 Introduction
2 Proposed Strategy
3 Analysis
4 Illustrative Example
5 Conclusion
Main Idea:
• Predictors using decoupled models to predict neighbor states
• Event trigger that monitors actual state and state of local predictor
• Event: Communication of state values with neighbors
• Robust MPC controller ([CRZ01]) to account for inaccuracies
x̂ i,t = Aii x̂ i,t−1 + B i û i,t−1 ,if x i,t − (Aii x̂ i,t−1 + B i û i,t−1 ) ∈ Ei (x i,t )
(
x i,t − x̂ i,t ∈ Ei (x i,t ) ⇔ x i,t − x̂ i,t ∈ Eiabs ∧ x i,t − x̂ i,t ∈ Eirel (x i,t )
MPC Controller
Prediction model for i ∈ I1:M over r ∈ I0:N−1 :
N−1
li (ū i,t+r − K ii x̄ i,t+r − K ij x̄ [i]j,t+r ), li (v i ) = kv i k2Li
X X
Cost function: JN,i =
r =0 j∈Ni
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 10/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Algorithm
• Measure x i,t
• If t = 0: Initialization
• Solve Initialization Problem
• Exchange the sequence V̂ i,0
• Evaluate the event trigger
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 11/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
1 Introduction
2 Proposed Strategy
3 Analysis
4 Illustrative Example
5 Conclusion
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 12/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Aij Ejabs ,
M
W̄i0 =
j∈Ni
W̌j0 =F jj Ejabs ⊕ B j Vj
F jk Xk ⊕ (−B j K jk ) Ekabs
Mh i
⊕
k∈Nj
Initialization
JN,i X̌
ˇ , Ǔ
minimize ˇ
i,0 i,0
X̌
ˇ ,Ǔ
i,0
ˇ
i,0
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 14/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Convergence
The closed-loop evolves according to x t+1 = Fx t + Bv ∗t − K D e t .
Assumptions:
• F = A + BK is Schur.
• For P 0, Q̃ 0, L = diag (Li ), S e = diag (S ei ) and S x = diag (S xi ):
P −F T PF −Q̃ −S x F T PBK D −F T PB
∗ S −K TD B T PBK D
e
−K D B PB 0
T T
∗ ∗ L−B T PB
Theorem 2
The closed-loop converges to the origin with region of attraction X N .
Idea of Proof:
• Use Lyapunov-like function VN∗ (x , V̌ ) = x T Px + JN∗ (x , V̌ )
• e t ∈ E rel is bounded by relative part of event-generator
• LMI condition ensures decay of VN∗ (x , V̌ ) over time
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 15/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
1 Introduction
2 Proposed Strategy
3 Analysis
4 Illustrative Example
5 Conclusion
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 16/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Simulation Set-up
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 17/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Results
2
CMPC
DMPC
1 DeMPC 1
abs rel
Velocity
index subsystem
0 3
-1
2
-2
1
-0.1 -0.05 0 0.05 0.1 0 0.05 0.1 0.15 0.2 0.25 0.3
Angle t in s
5
1 2 3 4
Input force
-5
0 0.02 0.04 0.06 0.08 0.1 0.12
t in s
1 Introduction
2 Proposed Strategy
3 Analysis
4 Illustrative Example
5 Conclusion
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 19/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Conclusion
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 20/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Questions?
Email: berkel@eit.uni-kl.de
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 21/ 21
Introduction Strategy Analysis Illustrative Example Conclusion
Felix Berkel, Steven Liu Non-Iterative DMPC with Event-Triggered Communication 21/ 21