Networked Sensing, Estimation and Control Systems

Vijay Gupta, Richard M. Murray, Ling Shi and Bruno Sinopoli

This page is intended for use by authors of the book "Networked Sensing, Estimation and Control Systems".

Chapter Table

Preface

• Motivation for writing the book
• Intended audience and prerequisites

I. Introduction to Networked Control Systems

• Motivation and examples
• Big picture for networked control systems
• Control systems (summarize basic framework; refer to standard texts for details)
• Communication systems
  • Channel models
  • Hardware protocols: 802.11, Zigbee, Wireless HART, SPA100, CANbus, FieldBus, ...
  • Software protocols: TCP, UDP, CSMACA (introduce basic protocols and indicate features that will come up later)
• Information patterns
  • Talk about distributed information: centralized versus decentralized, distributed
• Summary of the development of NCS

II. State Estimation and Sensor Fusion

Overview of estimation Gaussian random processes Kalman filtering Information form and fusion

References Durrant-Whyte

III. Information Theory

• Entropy and mutual information
• Bode's integral and Kalman filtering (from info theory point of view)
• Quantization tools (Nair, ...; 2007 IEEE Proc, Jan)
• Remark (19 Dec 08): do we want to include anything on rate distortion theory? Not really required for information below, but perhaps we want to add some material that could use that?

IV. Jump Linear Markov Systems

• Definitions and fundamental results
• Effects of packet loss
• Effects of delay
• Extensions to Control

V. Packet-Based Estimation (focus on single sensor)

Problem setup and useful lemmas Expected value of covariance Probabilistic bounds Sending measurements versus estimates Markov models for packet loss (JLMS) Multi-channel and/or multi-sensor?

References Sinopoli et al (TAC 2004) Epstein et al (Automatica 2008), Shi et al (TBD, 2008) Hespanha, Gupta et al Examples Building systems (temperature, humidity, etc) - first order models Vehicle tracking/MVWT - second order models (linearized?)

VI. Packet-Based Control (focus on single actuator)

TCP vs UDP LQG extensions Effects of quantization (?) Actuation buffers Information patterns Remark: should we include something about "scheduling" results? Go back to early Brockett work, then Walsh, etc? Nonlinear extensions

References Schenato et al (IEEE Proc, 2007) Kumar Gupta Sinopoli (Allerton 2007) Liberzon (nonlinear) Examples Vehicle systems/MVWT/Alice

VII. Information Flow on Graphs

Review of graph theory Average consensus (including rates) Switching, time delays, ... Gossip algorithms, load balancing, ...

References Fax, Olfati-Saber Jadbabaie et al Boyd, Shah et al Upfal (load balancing) Examples: Leader election, counting number of nodes, etc Distributed accounting (ala Spanos) Load balancing in computing systems

VIII. Distributed Estimation

Centralized fusion (information filters) Hierarchical fusion (plus measurements vs estimates?) Decentralized Kalman filtering (full connectivity -> graph) Effects of packet loss

References Speyer, Chong, Durrant-Whyte Examples Distributed sensor fusion for Alice

IX. Distributed Control

Laplacian/Nyquist theory Spatially invariant systems Multi-channel information flows (?) Distributed receding horizon control? Effects of packet loss Witsenhausen and information patterns

References Bamieh, Lall, Dullerud, D'Andrea (spatially invariant) Qiu work on multi-channel bandwidth allocation Examples Formation control (spacecraft, platoons)

X. Cooperative Control (application chapter)

Motivation Notions of complexity Formation, coverage, rendezvous Active sensing - sensors adjust operation to provide necessary information Protocols? How does this material relate to the rest of the book?

References Bullo, Cortes, Martinez Egerstedt Examples RoboFlag subproblems (TBD)

XI. Efficient Computation and Communications (application chapter)

• Shock aborbers
• Measurements versus estimates
• Actuation buffers
• Event-based control (transmit when necessary)
• Trading communication for computation (eg, estimate error at receiver; transmit new measurement if large)

XII. Implementation (application chapter)

• Robustness (byzantine)
• Synchronization (Lamport)

XIII. Sensor Networks (application chapter)

Motivation Sensor scheduling/topology/power Event-driven systems (?) Sensor selection/MAC layer How does this material relate to the rest of the book?

References Shi, Johansson, Murray (sensor topology/energy)

XIV. Future Directions and Open Problems

• Robustness (byzantine)
• Design of networks (utility-based theory, Low/Kelly)

Additional notes

Hong Kong, 19 Dec 08 (RMM + LS)

• Do we want to say something about bandwidth allocation in Alice as an example? Where does it fit? Could be a running example or perhaps go in one of the application chapters?

Related Information

Other books

• Bullo, Cortes, Martinez: book on cooperative control, but with a very different emphasis than Reza's proposed book. An online copy of the book is available at http://coordinationbook.info.

• Mesbahi and Egerstedt: graph theoretic methods for multi-agent coordination.

• Olfati-Saber: Networked Multi-Agent Systems: Distributed Algorithms for Coordination and Information Fusion

• D. Hristu-Varsakelis and W. S. Levine (Ed.): Handbook of Networked and Embedded Control Systems, 2005. ISBN 0-8176-3239-5
  • Mainly a collection of articles

Other information resources

• http://filer.case.edu/org/ncs

Upcoming events

Upcoming courses

• Apr 09: EECI Networked Control Systems course (Richard, Vijay)
• Fall 2009: NCS at HKUST (Ling)
• Spring 2010: NCS at CMU (Bruno)

Author meetings

• 25 Nov 08: Richard and Bruno in Pasadena
• 8-10 Dec 08: Bruno and Vijay at CDC
• 15 Dec 08: Richard and Ling in Hong Kong + Vietnam
• 13-19 Apr 09: HSCC, CPS week in San Francisco
• 11-17 Jun 09: Ling in Pasadena (commencement on 12 Jun)
• 27-29 Aug 09: Asian Control Conference (ASCC) in Hong Kong [Ling, Bruno, Richard]