Networked Control SystemsNetworked Control Systems

Vincenzo Liberatore

Today: CyberspaceToday: Cyberspace

Interact with remote virtual environment– On-line social activities

Communicate with co-workers, friends

– On-line search and information gathering– Disaster Recovery Networks– Distributed multimedia, games– E-commerce:

Buy books, CDs, computers Dump stock shares

Tomorrow: Cyber+PhysicsTomorrow: Cyber+Physics

Affect a remote physical environment– Hazardous environments– Terrestrial, space exploration– Distributed manufacturing– Remote diagnostic and troubleshooting– Experimental facilities

E.g., microgravity fluid dynamics

– Home robotics

Enabling TechnologiesEnabling Technologies

Embedded computing– Embed computers in the physical world

Pervasive networking– Interconnect them

Control theory and robotics– Physics-software interface

ChallengesChallenges

Remove dependency on– Precise environment models– Quality-of-Service provisioning

Software that is– Distributed– Adaptive and evolvable– Secure and safe (e.g., stability)– Remotely re-programmable

OutlineOutline

Remote robotic manipulationFoundations for Networked Control

Systems

Remote Robotic ManipulationRemote Robotic Manipulation

ObjectivesObjectives

Manipulation tasks– E.g., collect and manipulate a specimen from Mars

surface– Contact operations

Energy exchanged between robot and environment Results in physical change in environment state

– Safety: Force or kinematics constraints

Remote supervision

Distributed ControlDistributed Control

Natural admittance control– Virtual Attractors– Locally encapsulates real-time constraints

Multi-agent distributed system– Software components, autonomous, adaptable,

knowledgeable, mobile, collaborative, persistent

Supervisory control and tele-operation Evolvable software

DemonstrationDemonstration

Engineer, using a C# interface, loads a high-level command into the Virtual Supervisor (VS).

Wide-area Network

The VS then splits the command into pieces and send them to the

VR

VR

VR VR

VRVR

VR

Based on data obtained by the other VR’s, the robot is controlled directly by a single VR.

Virtual Robots (VR’s) residing on the network.

BenefitsBenefits

Supervisory control– Beyond tele-operation and autonomy

Dynamic reconfiguration– Task-oriented reconfiguration– Plug-and-play– Fault-tolerance– Survivability– Extensibility

Foundations ofFoundations ofNetworked Control SystemsNetworked Control Systems

Networked Control SystemsNetworked Control Systems

Traditional view

CONTROL SYSTEM+

-

Networked control system

INTERNET

CONTROL SYSTEM+

-

FoundationsFoundations

Toward a theory of networked control– Mathematical underpinnings – Established operation principles– Patterns and pitfalls

Core factors– System dynamics

E.g., output of open loop system

– Network dynamics E.g., packet forwarding, losses, delays, jitter

Networked control systems

AcknowledgmentsAcknowledgments

Faculty– Vincenzo Liberatore, Wyatt S. Newman,

Michael S. Branicky, Steve P. PhillipsStudents

– Adam Covitch, Ahmad Al-HammouriAlumni

– David Rosas

More info?More info?

Demos– Internet Robotics– Scalable Data Dissemination

URL

http://vorlon.cwru.edu/~vxl11/netlab/