distributed manufacturing control with extended cnp ...€¦ · distributed manufacturing control...
TRANSCRIPT
Distributed Manufacturing Control with Extended CNP Interaction
of Intelligent Products
Theodor Borangiu1, Silviu Raileanu1, Damien Trentesaux2, Thierry Berger2 , Iulia Iacob1
1 University Politehnica of Bucharest, Dept. of Automation and Industrial Informatics, ROMANIA
CIMR Centre of Research & Training in Robotics and CIM, [email protected] 2 Université Lille Nord de France, F-59000 Lille, UVHC, TEMPO Lab. F-59313 Valenciennes,
FRANCE
14th IFAC Symposium on Information Control Problems in Manufacturing
Bucharest, May 23-25, 2012
2
Summary
1. Introduction
2. Structure of the control model
3. Extended CNP for order holon execution
4. Information architecture
INCOM 2012, Bucharest, May 23-25, 2012
3
Introduction
Manufacturing domain is suitable for decentralization of the control because it features physically highly distributed processing systems
Production environment is not always deterministic, centralized approaches rapidly become inefficient when the shop floor cell must deal with disturbances or uncertainties => switch the primary objective of a designed system from global optimization to adaptability at perturbations and real-time optimization
Intelligence is highly distributed in sensing and local analysis of physical properties
Distributed intelligence paradigms: Product-driven automation Intelligent product (McFarlane et al., 2002, Meyer et al., 2008) Multi Agent control (Bellifemine et al., 2007) Holonic Manufacturing Systems (Babiceanu et al., 2004, Borangiu et al., 2009) Result: dynamic reconfiguration to provide agility to frequent changes in
production, fault-tolerance to resource breakdowns and adaptability to material flow variations.
Reduce the myopia (Leitao et al., 2010) through a combination of the two control modes - hierarchical and heterarchical
INCOM 2012, Bucharest, May 23-25, 2012
4
Structure of the control model
Passive
product
Product blueprint/
Product holon
Resources
Transportation
resources
Processing
resourcesStorage
resources
Space
transformation
Structure
transformation
Time
transformation
Intelligent products/
Order holons
(level 1)
Infrastructure
(routing)Pallet
AssociativeNon
associative
composes
composes
Intelligent
Embedded Device
composes
Resource holons
(level 1)
Is a
Supervisory entity
Coordinator holon
(level 2)
InfluenceInfluence
composes
Information
processing
Fabrication
machines
Material
processing
Is a
Is a
Dialog
Is a
Information
storageInformation
transportation
Is a
INCOM 2012, Bucharest, May 23-25, 2012
5
Structure of the control model
Order holon represent the client’s orders composed of: product, pallet carrier and IED
Intelligence boundary
Ph
ysi
ca
l p
ro
du
ct
Inte
rn
al
inte
ra
cti
on
s (a
sso
cia
tio
n,
dir
ect
co
ntr
ol)
Ex
tern
al
inte
ra
cti
on
s
(co
mm
un
ica
tio
n, se
nso
rs)
Memorization module
Communication
module
System model
IED
Processing
module
Product model
Info
rm
ati
on
al
en
vir
on
men
t
INCOM 2012, Bucharest, May 23-25, 2012
Structure of the control model
The proposed structure of the control system is completed with two additional elements which help ease the access to information Resource Service Access Model (RSAM):
distributed fault-tolerant entity in charge of collecting resource information during their usage and offering it in a concise manner when taking the decision of resource allocation.
Mediator: agent in charge with conflict resolution during active OH bidding.
6INCOM 2012, Bucharest, May 23-25, 2012
Extended CNP for order holon execution
Semi-heterarchical strategy
When executing a product the decisional module shouldprovide an answer to the following questions: "What is thenext operation to be performed? What is the resource thatwill do that operation? How do I bring the product there(which is the route)?"
7INCOM 2012, Bucharest, May 23-25, 2012
Strategy Description Expected results
Hierarchical Planning and scheduling
for a batch are centralized
done at the high layer
Minimize the
makespan at batch
level
Negotiated
Heterarchical
Schedules of each active
OH are computed through
communication with the
other active OHs.
Minimize the
makespan at
packet level
Non-Negotiated
Heterarchical
There is no global
scheduling and the next
operation is selected on
the first free resource
found
Minimize the
perturbation
impact
Knowledge
update
RSAM Mediator
Decision Execution
Resource
Holons
8
RSAM Mediator OHs RHs
Update HR
list
Choose best HRs
m
n≤mdeadline
refuse
accept
i≤n
j<n-i
reject
proposal
accept
proposal
k=j-1
1
Incre
ase R
H lo
ad
Delay
(due to
physical
constraints )Request op.start
Iterative
coordination
cfp
Request
Inform
failure
inform -done
inform -result
Next operations
through
No
rm
al o
pe
ra
tio
n m
od
e
Pe
rtu
rb
atio
n d
ete
cted
Extended CNP for order holon execution
INCOM 2012, Bucharest, May 23-25, 2012
9
Generic control model: based on
JADE framework
Composing agents
• Active holon entity agent
(Overo air)
• Resource holon agent (legacy
equipment integration through
MAS technology)
• RSAM agent
Host for resource j
PC
(X86 Architecture)
Host for order i
Embedded real-
time system (ARM
Architecture)
Wi-Fi / ETH Network
Host configuration
terminal
`
RS
AM
Ag
en
t
(in
itia
l
co
nfi
gu
rati
on
)
AH
E &
Me
dia
tor
Ag
en
t
RH
Ag
en
t
RS
AM
Ag
en
t
Server
(X86 Architecture)
JADE Distributed Agent Platform for FMS Control
JADE Main
Container
JADE Agent
Container
JADE Agent
Container
Resource controllerPhysical pallet
with productPhysical resource
Generic
VM
Generic
VMGeneric
VM
ETH connection
Electrical connection
Physical
association
RS
AM
Ag
en
t
Information architecture
INCOM 2012, Bucharest, May 23-25, 2012
10
Main RSAM
Agent
RH Agent
Services Search
Agent (Directory
Facilitator (DF))
Backup RSAM
AgentsBackup RSAM
AgentsBackup RSAM
Agents
(1) Request agent address
based on service type
(2) Inform
(3) Update HR status
(4) ACK status update
OH Agent
Resource information
collection/update process
(3) Request
service
(4) Provide
service
OH service reception
process
Information backup
process
Post
RSAM
service
Post RSAM service
(1) Request
RSAM
service
(2) Provide
service
Periodic Inform
RSAM High Availability Architecture
Information architecture
INCOM 2012, Bucharest, May 23-25, 2012
Experiments have been done using the three proposed control strategies
In a non-perturbed environment the hierarchical strategy offers the best production time (global optimization), followed by the negotiated heterarchical(local optimization) and finally by the non-negotiated heterarchical (no optimization).
In a perturbed environment (resource failures (flash) the results prove that the hierarchical strategy becomes inefficient compared with a heterarchical(negotiated of non-negotiated) strategy
11INCOM 2012, Bucharest, May 23-25, 2012
Conclusions
0
1000
2000
3000
4000
5000
6000
`10` `20` `30` `40` `50`
Pro
du
cti
on
du
rati
on
[s]
Number of products
Hierarchic
Negotiatedheterarchic
Non-Negotiatedheterarchic
0
100
200
300
400
500
600
700
800
900
1000
`1` `2` `3` `4` `5` `6` `7` `8` `9` `10` `11` `12`
Pro
du
cti
on
du
rati
on
[s]
Number of products
Hierarchic
Negotiatedheterarchic
Non-Negotiatedheterarchic
Advantages of using a semi-heterarchical strategy whose scope is to switch between hierarchical (stable FMS) and heterarchical (frequent disturbances) control.
12INCOM 2012, Bucharest, May 23-25, 2012
Conclusions