Download presentation
Presentation is loading. Please wait.
Published byAlfred Brown Modified over 8 years ago
1
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator (EVPS) group seminar on Oct. 19th, 2000 [presenter] Rong Zhang [advisor] Prof.. Andrew Alleyne [team partner] Eko Prasetiawan [project sponsor] Caterpillar ALLEYNE RESEARCH GROUP
2
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 2 Overview 1. Problem statement 2. Introduction to LQG/LTR control 3. EVPS LQG/LTR design 4. EVPS LQG/LTR performance 5. Conclusions
3
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 3 1. Problem statement Introduction to the Earthmoving Vehicle Powertrain An analogy between passenger vehicle powertrain and EVP EVPS schematic and I/O list Need for coordination A tracking example 1. Problem statement 2. Introduction to LQG/LTR control 3. EVPS LQG/LTR design 4. EVPS LQG/LTR performance 5. Conclusions
4
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 4 An analogy Passenger Vehicle Prime mover: Usually Spark-Ignition type engine (gas) Torque Converter: Mechanical gearbox Resistance speed control: Brake Earthmoving Vehicle Prime mover: Usually Compression-Ignition type engine (diesel) Torque pressure converter: Hydraulic pump Resistance speed control: Flow valve
5
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 5 3 4 5 1 2 1 23 EVPS schematic...
6
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 6 … and I/O list A MIMO control system Controlled outputs: load speeds ( 3)
7
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 7 Need for coordination! A tracking example Tracking references Node 1: A rising step Node 2: 0 Node 3: 0 Using only one input: Flow Valve 1... n m1 (rpm) n m2 (rpm) n m3 (rpm )
8
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 8 5 control inputs without coordination... [Q] How to take actions at the right time, right direction and right amount? [A] Coordination needed !
9
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 9 2. Introduction to LQG/LTR control Pole placement? “Performance” vs. “Cost” LQR controller -- “optimal” feedback law LQR estimator -- “optimal” filter by Kalman LQG controller design Optimal controller + Optimal estimator LQG/LTR controller design Optimal + Optimal Robustness 1. Problem statement 2. Introduction to LQG/LTR control 3. EVPS LQG/LTR design 4. EVPS LQG/LTR performance 5. Conclusions
10
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 10 Pole placement? [Q1] Where should the target poles be placed? Too slow? poor performance! Too fast? expensive controller and surprising power bill! [Q2] Is there an “optimal” controller balancing both Performance and Cost? “Punishment philosophy” Poles will be here!
11
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 11 LQR controller In this method, pole locations are not designed directly. Instead, find a good u=-Kx that minimizes: Q and R are Performance Index or “Punishment Matrices” Want a quicker state convergence? make Q bigger to punish large states! Want to keep control efforts within saturation range or at a lower cost? make R bigger to punish overacting inputs! [Solution] Theoretical: ARE equation finds us a good K Practical: Matlab command ‘lqr’
12
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 12 LQR estimator Not all the states are available, how to construct them from y? An estimator Find a good L(u e =-LCe) that minimizes: If Q e and R e are determined by process and measurement noise level... A Kalman Filter! [Solution] Theoretical: ARE equation Practical: Matlab command ‘lqr’ ‘kalman’
13
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 13 LQG = LQR control + Kalman filter LQG/LTR = LQG + Robustness recovery It’s a Optimal + Optimal design, but is it “optimal” in the sense of robustness? No! Using a recovery procedure (r=0 to inf), to make the LQG closed-loop closer to that of the Target Loop: the ideal LQR loop with full-state feedback. [Solution] Theoretical: Loop Transfer Recovery procedure Practical: Matlab command ‘ltru’ ‘ltry’
14
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 14 Optimal + Optimal Robustness LQR with Full-state feedback Singular Value Bode Plot Singular Value Bode Plot LQG with measurements feedback r = 0 (no recovery)
15
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 15 Loop transfer recovery... r = 1 (small recovery)r = 10 5 (large recovery) Closer to the target loop Singular Value Bode Plot Singular Value Bode Plot
16
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 16 3. EVPS LQG/LTR design Plant Model (14 states) to Design Plant Model (17 states) To insure 0 tracking errors to step inputs, the PM is augmented by 3 free integrators. LQG design Good “Punishment Matrices” are found and tested LQG/LTR design Robustness or the ideal LQR is recovered 1. Problem statement 2. Introduction to LQG/LTR control 3. EVPS LQG/LTR design 4. EVPS LQG/LTR performance 5. Conclusions
17
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 17 EVPS System Plant Model A 14x14 Design Plant Model A 17x17 Three 1/s’ added to insure 0 tracking error
18
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 18 LQG Controller LQG/LTR Controller
19
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 19 4. EVPS LQG/LTR performance Simultaneous tracking Different nodes track different speed references The total flow demand changes Disturbance rejection One of the 3 nodes is subject to a pressure disturbance The TOTAL flow demand does not change The distribution of pressures among the 3 nodes is changed 1. Problem statement 2. Introduction to LQG/LTR control 3. EVPS LQG/LTR design 4. EVPS LQG/LTR performance 5. Conclusions
20
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 20 Simultaneous speed-tracking n m1 tracking +/- 100rpm reference n m2 being regulated n m3 tracking - 60rpm reference n m1 (rpm) n m2 (rpm) n m3 (rpm ) Opposite directionSame direction
21
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 21 Pressures of simu-tracking p d1 increased to push through more flow p d2 unchanged to maintain the same flow p d3 decreases to push through less flow p d1 (MPa) p d2 (MPa) p d3 (MPa)
22
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 22 Control inputs of simu-tracking Throttle when total flow demand Pump when total flow demand Flow 1 when speed reference 1 Flow 2 compensates for pressure resulted from total flow Flow 3 when speed reference 3
23
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 23 Pressure disturbance at node 1 Pressure step as disturbance is applied at node 1 only p d1 (MPa) p d2 (MPa) p d3 (MPa) Neighbor node pressure doesn’t change significantly
24
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 24 Speeds of disturb. rejection n m1 (rpm) n m2 (rpm) n m3 (rpm ) n m1 decreases when disturbance pressure squeezes out some flow; then regulated by the controller n m2 increases by pressure disturbance squeezes in some flow from neighbor node; then regulated by the controller n m3 increases by pressure disturbance squeezes in some flow from neighbor node; then regulated by the controller
25
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 25 Control inputs of disturb. rejection Throttle compensates for small total pressure Pump doesn’t need to change much Flow 1 to fight disturbance pressure Flow 2 compensates for upstream pressure caused by load 1 Flow 3 compensates for upstream pressure caused by load 1 total flow demand not changed!
26
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 26 5. Conclusions An LQG/LTR MIMO controller is successfully designed and implemented The system: 14 states, 9 measurements, 5 inputs The design plant model with free integrators: 17 states The LQG/LTR controller: 17 states, 9 inputs, 5 outputs It has satisfying tracking and disturbance rejecting performance It’s robustness and working range are subject to further validation Model reduction technique will be used to simplify the controller 1. Problem statement 2. Introduction to LQG/LTR control 3. EVPS LQG/LTR design 4. EVPS LQG/LTR performance 5. Conclusions
27
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 27 References M. Athans, "A tutorial on the LQG/LTR method," presented at American Control Conference, Seattle, WA, 1986. A quick start. B. D. O. Anderson and J. B. Moore, Optimal Control, Linear Quadratic Methods. Eaglewood Cliffs, New Jersey: Prentice-Hall, 1990. A textbook. J. C. Doyle and G. Stein, "Multivariable Feedback Design: Concepts for a Classical/Modern Synthesis," IEEE Trans. Automat. Contr., vol. AC-26, pp. 4-16, 1982. A classic. A. Saberi, B. M. Chen, and P. Sannuti, Loop Transfer Recovery: Analysis and Design. London: Springer-Verlag, 1993. A monograph. Matlab manual online “Robust Control Toolbox” at: http://www.mathworks.com/access/helpdesk/help/pdf_doc/robust/ robust.pdf A useful tool.
28
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang ALLEYNE RESEARCH GROUP, M&IE/UIUC 28 An earthmoving vehicle powertrain Drive Hydr. Pump Hydr. Pump Steering Implement Engine 5 4 3 21 Control
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.