Model Predictive Control Jn Drgoa (STU) ECC16 June 30, 2016 2 / - PowerPoint PPT Presentation

Regionless Explicit MPC of a Distillation Column Ján Drgoňa , Filip Janeček, Martin Klaučo, and Michal Kvasnica Slovak University of Technology in Bratislava, Slovakia June 30, 2016 I AM Ján Drgoňa (STU) ECC16 June 30, 2016 1 / 17

Model Predictive Control Ján Drgoňa (STU) ECC16 June 30, 2016 2 / 17

Implicit vs Explicit MPC Explicit MPC Implicit MPC Ján Drgoňa (STU) ECC16 June 30, 2016 3 / 17

Implicit vs Explicit MPC Explicit MPC Implicit MPC Large systems Expensive implementation Matrix inversions Harder to certify Ján Drgoňa (STU) ECC16 June 30, 2016 3 / 17

Implicit vs Explicit MPC Explicit MPC Implicit MPC Large systems Small systems Expensive implementation Cheap implementation Matrix inversions Division-free Harder to certify Rigorous analysis Ján Drgoňa (STU) ECC16 June 30, 2016 3 / 17

Implicit vs Explicit MPC Explicit MPC Implicit MPC Large systems Small systems Expensive implementation Cheap implementation Matrix inversions Division-free Harder to certify Rigorous analysis Regionless Explicit MPC Ján Drgoňa (STU) ECC16 June 30, 2016 3 / 17

Region-Based Explicit MPC 1 2 U T HU + θ T FU min U s.t. GU ≤ w + S θ Ján Drgoňa (STU) ECC16 June 30, 2016 4 / 17

Region-Based Explicit MPC U ⋆ = V i θ + v i λ ⋆ = Q i θ + q i R i = { θ | GU ⋆ ≤ w + S θ, λ ⋆ ≥ 0 } Ján Drgoňa (STU) ECC16 June 30, 2016 4 / 17

Active Sets – Geometric Method 1 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

Active Sets – Geometric Method 1 Pick an arbitrary point in Ω . 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

Active Sets – Geometric Method 1 Construct an initial critical region. R 1 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

Active Sets – Geometric Method 1 Pick new points in Ω . R 1 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

Active Sets – Geometric Method 1 R 3 R 2 R 4 Construct new critical regions. R 5 R 1 R 7 R 6 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

Active Sets – Geometric Method 1 R 3 R 2 R 4 Repeat recursively until whole Ω is R 5 R 1 covered. R 7 R 6 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

Active Sets – Geometric Method 1 R 11 R 10 R 9 R 12 R 3 R 2 R 4 R 13 Repeat recursively until whole Ω is R 8 R 5 R 1 covered. R 14 R 7 R 19 R 15 R 6 R 17 R 18 R 16 1 M. Baotić: Optimal Control of Piecewise Affine Systems – A Multi-parametric Approach, 2005 Ján Drgoňa (STU) ECC16 June 30, 2016 5 / 17

The Idea R 11 R 10 R 9 R 12 R 3 R 2 R 4 R 13 Avoid the construction and the R 8 R 5 R 1 R 14 storage of regions! R 7 R 19 R 15 R 6 R 17 R 18 R 16 Ján Drgoňa (STU) ECC16 June 30, 2016 6 / 17

The Idea R 11 R 10 R 9 R 12 R 3 R 2 R 4 R 13 U ⋆ = V i θ + v i V ( A i ) R 8 R 5 R 1 λ ⋆ = Q i θ + q i Q ( A i ) R 14 R 7 R 19 R 15 R 6 R i = { θ | GU ⋆ ≤ w + S θ, λ ⋆ ≥ 0 } R 17 R 18 R 16 Ján Drgoňa (STU) ECC16 June 30, 2016 6 / 17

The Idea 2 A 11 A 10 A 9 A 12 A 3 A 2 A 4 A 13 U ⋆ = V ( A i ) θ + v ( A i ) A 8 A 5 A 1 λ ⋆ = Q ( A i ) θ + q ( A i ) A 14 A 7 A 19 A 15 A 6 A = {A 1 , . . . , A R } A 17 A 18 A 16 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 6 / 17

Active Sets – Extensive Enumeration 3 A 11 A 10 {} A 9 A 12 A 3 A 2 A 4 A 13 { r } { 1 } { 2 } { 3 } . . . A 8 A 5 A 1 A 14 A 7 A 19 A 15 A 6 { 1 , 2 } { 1 , 3 } { 1 , r } { 3 , 1 } { 3 , 2 } { 3 , r } . . . . . . A 17 A 18 A 16 { 1 , 3 , 2 } { 1 , 3 , r } { 3 , 1 , 2 } { 3 , 1 , r } . . . . . . 3 A. Gupta et. al.: A novel approach to multiparametric quadratic programming, Automatica 2011 Ján Drgoňa (STU) ECC16 June 30, 2016 7 / 17

Active Sets – Extensive Enumeration 3 – Tree Pruning A 11 A 10 {} A 9 A 12 A 3 A 2 A 4 A 13 { r } { 1 } { 2 } { 3 } . . . A 8 A 5 A 1 A 14 A 7 A 19 A 15 A 6 { 1 , 2 } { 1 , 3 } { 1 , r } { 3 , 1 } { 3 , 2 } { 3 , r } . . . . . . A 17 A 18 A 16 { 1 , 3 , 2 } { 1 , 3 , r } { 3 , 1 , 2 } { 3 , 1 , r } . . . . . . 3 A. Gupta et. al.: A novel approach to multiparametric quadratic programming, Automatica 2011 Ján Drgoňa (STU) ECC16 June 30, 2016 7 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase On-line phase Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase Active Sets 1 On-line phase Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase Active Sets 1 Geometric method On-line phase Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase Active Sets 1 Geometric method Extensive enumeration On-line phase Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase Active Sets 1 Geometric method Extensive enumeration U ⋆ , λ ⋆ from KKT conditions 2 On-line phase Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase Active Sets 1 Geometric method Extensive enumeration U ⋆ , λ ⋆ from KKT conditions 2 Construction of regions 3 On-line phase Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Region-Based Approach Off-line phase Active Sets 1 Geometric method Extensive enumeration U ⋆ , λ ⋆ from KKT conditions 2 Construction of regions 3 On-line phase Function evaluation 4 Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Explicit MPC Solution Recap Regionless Approach Off-line phase Active Sets 1 Geometric method Extensive enumeration U ⋆ , λ ⋆ from KKT conditions 2 Construction of regions 3 On-line phase Point location 4 Ján Drgoňa (STU) ECC16 June 30, 2016 8 / 17

Point Location Problem 2 A 11 A 10 A 9 A 12 A 3 A 2 ? A 4 A 13 How to find in which region we are, A 8 A 5 A 1 without storing any regions? A 14 A 7 A 19 A 15 A 6 A 17 A 18 A 16 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 9 / 17

Point Location Problem 2 A 11 A 10 A 9 A 12 A 3 A 2 ? A 4 A 13 U ⋆ = V ( A i ) θ + v ( A i ) A 8 A 5 A 1 λ ⋆ = Q ( A i ) θ + q ( A i ) A 14 A 7 A 19 A 15 A 6 R ( A i ) = { θ | GU ⋆ ≤ w + S θ, λ ⋆ ≥ 0 } A 17 A 18 A 16 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 9 / 17

Point Location Problem 2 A 11 A 10 A 9 A 12 A 3 A 2 ? A 4 A 13 U ⋆ = V ( A 1 ) θ + v ( A 1 ) A 8 A 5 A 1 λ ⋆ = Q ( A 1 ) θ + q ( A 1 ) A 14 A 7 A 19 A 15 A 6 R ( A 1 ) = { θ | GU ⋆ ≤ w + S θ, λ ⋆ ≥ 0 } A 17 A 18 A 16 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 9 / 17

Point Location Problem 2 A 11 A 10 A 9 A 12 A 3 A 2 ? A 4 A 13 U ⋆ = V ( A 2 ) θ + v ( A 2 ) A 8 A 5 A 1 λ ⋆ = Q ( A 2 ) θ + q ( A 2 ) A 14 A 7 A 19 A 15 A 6 R ( A 2 ) = { θ | GU ⋆ ≤ w + S θ, λ ⋆ ≥ 0 } A 17 A 18 A 16 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 9 / 17

Point Location Problem 2 A 11 A 10 A 9 A 12 A 3 U ⋆ A 2 A 4 A 13 U ⋆ = V ( A 2 ) θ + v ( A 2 ) A 8 A 5 A 1 λ ⋆ = Q ( A 2 ) θ + q ( A 2 ) A 14 A 7 A 19 A 15 A 6 R ( A 2 ) = { θ | GU ⋆ ≤ w + S θ, λ ⋆ ≥ 0 } A 17 A 18 A 16 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 9 / 17

Sequential search 2 1: procedure Regionless ( θ ) for i ∈ { 1 , . . . , R } do 2: Compute λ = Q ( A i ) θ + q ( A i ) 3: if λ ≥ 0 then 4: Compute U = V ( A i ) θ + v ( A i ) 5: if GU ≤ w + S θ then 6: U ⋆ ← U 7: return U ⋆ 8: end if 9: end if 10: end for 11: 12: end procedure 2 F. Borrelli and M. Baotić: On the computation of linear model predictive control laws, Automatica 2008 Ján Drgoňa (STU) ECC16 June 30, 2016 10 / 17

Model Predictive Control Jn Drgoa (STU) ECC16 June 30, 2016 2 / - PowerPoint PPT Presentation

Regionless Explicit MPC of a Distillation Column Jn Drgoa , Filip Janeek, Martin Klauo, and Michal Kvasnica Slovak University of Technology in Bratislava, Slovakia June 30, 2016 I AM Jn Drgoa (STU) ECC16 June 30, 2016 1 / 17 Model

Generalized Model Predictive Control (Discretely Generalized MPC) Sa sa V. Rakovi c, Ph.D.

Predicting the Future of Model Predictive Control Manfred Morari In Honor of Professor David

to our dear friend : David CLARKE ! .Defence : How to control a process with a non stationary

Model Predictive Control Model Predictive Control of Hybrid Systems of Hybrid Systems Model

MODEL PREDICTIVE CONTROL OF A 3D CRANE FOR EFFECTIVE TRACKING OF SQUARE AND STAR SHAPES

Title: Model reduction and thermal regulation by Model Predictive Control of a new cylindricity

Model Predictive Control: Background B. Wayne Bequette Concise Review of Undergraduate Process

Non-linear MPC Robert Platt Northeastern University NonLinear Model Predictive Control Given:

On Optimal Input Design in System Identification for Model Predictive Control Mariette

Multiple Model Predictive Control (MMPC) for Nonlinear Systems and Improved Disturbance Rejection

H0K03a : Advanced Process Control Model-based Predictive Control 1 : Introduction Bert Pluymers

H0K03a : Advanced Process Control Model-based Predictive Control 2 : Dynamic Optimization Bert

Flood control of river systems with Model Predictive Control The river Demer as case study

Introduction to Model Predictive Control (MPC) Oscar Mauricio Agudelo Maozca Bart De Moor

Introduction to Model Predictive Control (MPC) Oscar Mauricio Agudelo Maozca Bart De Moor

Efficient Symbolical and Numerical Algorithms for nonlinear model predictive control with

H0K03a : Advanced Process Control Model-based Predictive Control 3 : Stability Bert Pluymers Prof.

H0K03a : Advanced Process Control Model-based Predictive Control 4 : Robustness Bert Pluymers Prof.

A Step Beyond the State of the Art Robust Model Predictive Control Synthesis Methods by c

Outline n From linear to nonlinear n Model-predictive control (MPC) n POMDPs Page 1

Robustness of model-based control Emo Todorov Roboti LLC University of Washington Model-based

Reinforcement Learning and Model Predictive Control RL : optimizes policy for given cost

Pollution-based model predictive control of combined sewer networks, considering uncertainty

A Distributed Algorithm for Scenario-based Model Predictive Control using Primal Decomposition