Collaborative Agent Systems for Distributed Automation - CASDA -

1
(No Transcript)
2
Collaborative Intelligent SystemsWilliam A.
GruverDilip B. KotakFLINT CIBI 2003UC
Berkeley - 15 December 2003
3
CIS Structure
Monitoring Identification
Communi- cations
Collaboration Coordination
Systems Architecture
Manufac. Supply Chain
Service
Infrastructure Elec. Power
Infrastructure Transport.
4
Technologies

• Self Monitoring, Identification Control
• Hardware sensors, monitoring
  actuators Auto-ID Software Soft Computing
• Agent Communications Infrastructure
• Hardware wired, wireless RF, 802.11
• Software ontology, protocols, security FIPA,
  JADE
• Agent Collaboration Coordination
• Negotiation Contract Net, Blackboard, Auctions
• Soft Computing Fuzzy Logic, Neural Networks,
  Genetic Algorithms
• Agent Architecture
• Hardware Distributed Computing
• Software PROSA, Multi-Agent Architectures

5
Applications

• Manufacturing Supply Chain Sector
• Wood Products Manufacturing Loewen Windows
• Service Sector
• Health, Banking, Insurance, Food
• Transportation Infrastructure
• Public, Fleet
• Energy Infrastructure
• Electrical Distributed Power
• Hydrogen Alternative Fuels Mobile
  Stationary Fuel Cells

6
Typical Rough Mill Layout
Chop Saw
Scanner
Rip Saw
Scanner
LUMBER WAREHOUSE
Lumber
7
Summary of Key Issues

• Select most suitable jag 2,500 choices 30-50
  times / day
• Assign arbor/ripping priority 18 choices 5-10
  times / day
• Schedule components on kickers 5 x 1020
  choices 150 times / day
• Coordinate these three
• Optimizing the overall performance of the
  following
• Costs
• Yield
• Productivity
• Grade length utilization
• Order-file satisfaction
• On time delivery

8
System Architecture - for Rough Mill Production
Local optimization
Rip each piece of lumber into a best combination
of strip by width
Agent 1
Agent 2
Mediator Agent
Select a best load or jag of lumber from the
warehouse

• Facilitators functions
• Static optimized overall schedule
• Task decomposition and allocation
• System coordination and decision making
• Learning

Agent 3
Optimize component schedule to produce a best
combination of components for each strip
9
Applications Three Layers

• Manufacturing
• Supply Chain Factory coordination of logistics
• Work-Cell coordination of production
• Device intelligent autonomous devices
• Energy Infrastructure
• Inter Community coordination of supply demand
• Intra-Community coordination of devices
• Devices
• Photovoltaic cell, wind turbine, electrolyzer,
  reformers, storage, compressor, dispensers
• Generalize to Multi-Layered Holarchy
• Hierarchy of collaborative agents

10
Intelligent Wireless MicroRouter

• Multi-point, multi-hop, wireless connectivity of
  computers, sensors, displays, PDAs
• Intelligent routing and priority scheduling over
  extended distances at 11-54 Mbs
• Based on IEEE 802.11x protocol standards
• Distributed systems implementation in JADE
• In development by Intelligent Robotics
  Corporation and Simon Fraser University

11
Extended Coverage using 802.11
12
Features

• Hubless
• Self organizing connectivity without centralized
  communication
• Intelligent
• Self configuring without server or external
  router
• Distributed scheduling of tasks by priority,
  time, due dates
• Wireless
• Protocol IEEE 802.11b/g
• Interfaces USB2.0, RS232, IEEE 802.3
• Scalable
• Additional network devices and services may be
  easily added
• Robust
• Adaptability to user demands and network failures
  
• Secure
• Advanced encryption algorithms

13
Industrial ApplicationUtility Metering

• Enables utility meters to be networked without
  the need for hubs
• Utility meter data intelligently routed over
  networks with millions of nodes
• Provides network management, load balancing, and
  priority scheduling of services
• Applications also to transportation and other
  infrastructure systems

14
Distributed Combinatorial Scheduler

• Process
• task performed by one agent
• Operation
• sequence of processes
• never revisits an agent
• Fully distributed algorithm
• Software
• multi-threaded C and OpenGL
• being ported to Java for operation in JADE
  environment

DCS2
DCS1
15
Conclusions

• Collaborative intelligent systems have broad
  applications
• manufacturing resource management, planning, and
  control
• transportation, energy, and other infrastructure
  systems
• service systems
• Collaborative intelligent systems provide
• improved flexibility
• reduced setup-time
• higher robustness
• integration of human intelligence
• Collaborative intelligent systems are
  characterized by
• reconfigurability
• task level programming
• high-level cooperation

16
Further information
www.ensc.sfu.ca/irms Intelligent Robotics and
Manufacturing Systems Laboratory www.ieeesmc.
org IEEE-SMC Technical Committee
Collaborative Intelligent Systems
hms.ifw.uni-hannover.de Holonic Manufacturing
Systems Consortium www.fipa.org
Foundation for Intelligent Physical
Agents www.ims.org Intelligent Manufacturing
Systems Program
17
(No Transcript)