... 即時控制系統設計 Design of Real-Time Control Systems. Real-Time Control Systems. Lecture 32 Introduction to Networked Control Systems.

Transcription

1 Introduction NTUEE-RTCS32-NCS-2 SPRING 2010 即時控制系統設計 Design of Real-Time Control Systems Lecture 32 Introduction to Networked Control Systems Real-Time Control Systems Controlled by one Computer Processor Centralized control systems Real-time operating systems Controlled by one Communication Medium Distributed control systems Real-time communications Feng-Li Lian NTU-EE Feb10 Jun10 Centralized Control System Distributed Control System 04/12/03 Introduction NTUEE-RTCS32-NCS-3 Introduction NTUEE-RTCS32-NCS-4 Networks: Enable remote data transfers & data exchange among users Reduce the complexity in wiring connections and the costs of media Provide ease in maintenance Data Networks: > Slotted ALOHA & ARPANET around > Ethernet around 1980 Control Networks: > CAN (Controller Area Network) in 1983:» By Robert Bosch, Germany, for car industries > PROFIBUS (PROcess FIeld BUS) in 1987:» By six German companies & five German institutes > DeviceNet in 1994 (?):» By Allen-Bradley/Rockwell Automation, for manufacturing-related industries Networked Control Systems: Control systems with physically distributed processing power and network communication of control signals Information/System Network Discrete-Event/Cell Network Continuous-Variable/Device Network S A C Machine Tool... S A C S A C actuator delays Automatic Material Assembly Line Handling network channel... actuator input u1 u2 um vm v2 v1 plant controller y1 y2 yr reference input wr w2 w1 sensor output sensor delays

2 Introduction NTUEE-RTCS32-NCS-5 NTUEE-RTCS32-NCS-6 Motivations: The overall NCS performance is always affected by network delays Research Overview: NCS Configuration» Direct structure Delays are widely known to degrade the performance of a control system» Hierarchical structure Delays in-the-loop Existing constant time-delay control methodologies may not be directly suitable for controlling a system over the network since network delays are usually time-varying, especially in the Internet Therefore, to handle network delays in a closed-loop control system over a network, an advanced methodology is required Delay Characteristics» Cyclic service network» Random access network Effect of Delays in-the-loop > Performance degradation > Destabilization NTUEE-RTCS32-NCS-7 NTUEE-RTCS32-NCS-8 Direct structure: Direct structure: Hierarchical structure: Overstreet & Tzes 99

3 NTUEE-RTCS32-NCS-9 NTUEE-RTCS32-NCS-10 Direct structure: Direct structure: Overstreet & Tzes 99 Overstreet & Tzes 99 NTUEE-RTCS32-NCS-11 NTUEE-RTCS32-NCS-12 Delays in-the-loop: Hierarchical structure: reference signal control signal output signal Tipsuwan & Chow 02

4 NTUEE-RTCS32-NCS-13 NTUEE-RTCS32-NCS-14 Delays in-the-loop: sc : sensor-to-controller delay Delays in-the-loop: Node A Application Data Link Physical Twait Twait Ttx Node B Application Data Link Physical Ttx ca : controller-to-actuator delay Total end-to-end delay is the sum of Pre-processing time: microprocessor Waiting time: network protocol - MAC Transmission time: data rate & length Post-processing time: microprocessor Depend on network protocol and loading NTUEE-RTCS32-NCS-15 NTUEE-RTCS32-NCS-16 Delay Characteristics: Cyclic Service Networks: IEEE 802.4, SAE token bus, PROFIBUS, IEEE 802.5, SAE token ring, MIL-STD-1553B, FIP Control and sensory signals are transmitted in a cyclic order with deterministic behaviors Delays are periodic & can be simply modeled as a periodic function such as sc k = sc k+n and ca k = ca k+m, N, M are constants In practice, NCS may experience small variations on periodic delays due to several reasons such as discrepancies in clock generators Delay Characteristics: Random Access Networks: Ethernet, CAN (?) Significant parts of random network delays are waiting time delay due to queuing and frame collision on the networks Sources of randomness: > the queuing time delay at a switch or a router > The propagation time delays from different network paths Delay models: > Constant delay > Poisson process such as Markov chain > Fluid flow model, ARMA model etc.

5 NTUEE-RTCS32-NCS-17 NTUEE-RTCS32-NCS-18 Delay Models: Delay Models: Luck & Ray 90, Krtolica et al. 94, Nilsson 98 Luck & Ray 90, Krtolica et al. 94, Chan & Ozguner 95, Nilsson 98 NTUEE-RTCS32-NCS-19 NTUEE-RTCS32-NCS-20 Delay Models: Delay Models: Nilsson 98, Krtolica et al. 94, Nilsson 98

6 NTUEE-RTCS32-NCS-21 NTUEE-RTCS32-NCS-22 Experimental Data of Network Delays: Experimental Data of Network Delays: Low load network traffic Period: 50 ms Data rate: 10 kbps Data size: 8 bytes/128 bits Tx time: 12.8 ms CPU time: 3.4 ms Nilsson 98 Nilsson 98 NTUEE-RTCS32-NCS-23 NTUEE-RTCS32-NCS-24 Experimental Data of Network Delays: Add two additional low-load nodes with different priorities Experimental Data of Network Delays: Nine-node network Master Slave1 Slave2 Slave9 Ttx Twait Twait Ttx Nilsson 98

7 Overview of NCS Research Issue NTUEE-RTCS32-NCS-25 Overview of NCS Research Issue NTUEE-RTCS32-NCS-26 Performance degradation: Performance degradation: Overview of NCS Research Issue NTUEE-RTCS32-NCS-27 Overview of NCS Research Issue NTUEE-RTCS32-NCS-28 Destabilization: Root locus approach Destabilization: Bode plot approach db M(w) H B E D K = 10 K = 2 M(w) (rad/sec) (w) w1 w2 w3 (w) h = 0 (rad/sec) Malek-Zavarei & Jamshidi 87

8 Overview of NCS Research Issue NTUEE-RTCS32-NCS-29 Networked Control Methodology NTUEE-RTCS32-NCS-30 Networked Control Methodology: Destabilization: Bode plot approach db M(w) 0 H B E D -20 K = 10 (rad/sec) M(w) -40 K = 2-60 h = 4 h = 1 h = F h = A (w) C h = 0.1 h = h = 0 (rad/sec) w1 w2 w3 (w) -270 Assumptions 1. Augmented Deterministic Discrete-Time Model Methodology 2. Queuing Methodology 3. Optimal Stochastic Control Methodology 4. Perturbation Methodology 5. Sampling Time Scheduling Methodology 6. Robust Control Methodology 7. Fuzzy Logic Modulation Methodology 8. Event-Based Methodology 9. End-User Control Adaptation Methodology Malek-Zavarei & Jamshidi 87