© 2004 ipd, prof. lockemann universität karlsruhe (th) sofsem04 flexibility through multiagent...
TRANSCRIPT
© 2004 IPD, Prof. Lockemann
UniversitätKarlsruhe (TH)
SOFSEM04
Flexibility ThroughMultiagent Systems:Solution or Illusion?
Peter C. Lockemann
© 2004 IPD, Prof. Lockemann
2
SOFSEM04
OEM
Dispoweb
KRASH-MAS
IntaPS-MAS
FABMAS-MAS
dispowebdispowebATT/SCC
intraplant interplant external
Mon
itorin
gExe
cutio
nPlann
ing
Customer
Global planning and negotiations
Production planning for electronics
Assembly planning
Production planning for mechanics
Tracking
Scenario: Production and supply chain
© 2004 IPD, Prof. Lockemann
3
SOFSEM04
KRASH-MAS
ATT-MAS
DISPOWEB-MAS
Tracing
Planing
IntaPS-MAS
ATT-MAS
ATT-MAS
DISPOWEB-MAS
FABMAS
DISPOWEB-MAS
Negotiation of the global production schedule
Tracing of processes and propagation of delays
SCC-MAS
PlaningExe-cutionPlaning
Tracing
Exe-cution
Tracing
Exe-cution
KRASH-MAS
ATT-MAS
DISPOWEB-MAS
Tracing
Planing
IntaPS-MAS
ATT-MAS
ATT-MAS
DISPOWEB-MAS
FABMAS
DISPOWEB-MAS
Negotiation of the global production schedule
Tracing of processes and propagation of delays
SCC-MAS
PlaningExe-cutionPlaning
Tracing
Exe-cution
Tracing
Exe-cution
Scenario: Production and supply chain
© 2004 IPD, Prof. Lockemann
4
SOFSEM04
Scenario: Production and supply chain
Nr. Activity (Actor)
Negotiate initial plan of production between supply chain partners (DISPOWEB).
Operational assembly planning (KRASH). Production planning for e.g. mechanical parts (IntaPS). Production planning for e.g. electronic parts (FABMAS).
Monitoring of orders and related suborders (ATT). Trigger internal planning systems in case of minor critical events (ATT). Next Trigger re-planning by DISPOWEB agents in case of a severe critical event (ATT). Next Controlling information is forwarded to trusted third party SCC-system (ATT/SCC).
Internal rescheduling in reaction to a critical event (KRASH, IntaPS, FABMAS). Next
Renegotiate plan of production between supply chain partners due to severe critical event (DISPOWEB). Next
© 2004 IPD, Prof. Lockemann
5
SOFSEM04
(Referring) Physicians
TU Ilmenau/Physicians Thuringia, OnkoNet
Ambulance
TU Ilmenau/Ambulances Thuringia
Reception Ward
Uni Hohenheim/Hosp. Sindelfingen
Kliniken & Stationen
Gastro-Enterology
Uni Würzburg/Univ.-Hosp. Würzburg
Cancerology
TU Ilmenau/Univ.-Hosp. KIM II Jena
Internal Medicine
Uni Hohenheim/Hosp. Sindelfinden
Surgery Ward
Uni Mannheim/Hosp. Pegnitz, Hosp. Kulmbach
N.N.
(…/…)
Intensive Care Unit
TU Berlin/Hosp. Charité Berlin
Cardiological Laboratory
Uni Mannheim/Hosp. Pegnitz & Hosp. Kulmbach
Transport Serv.
Uni Freiburg/Univ.-Hosp. Freiburg
Operating Theatre
Surgery Block
Univ Trier/Hosp.Barmh. Brüder Trier
Op.-Planning
TU Berlin/Charité Berlin
Dept. of Surgery
Uni Mannheim/Hosp. Pegnitz & Hosp. Kulmbach
Emergency Dept.
TU Berlin/Hosp. Charité Berlin
Radiology & Radiotherapy
Conv., CT, MRT
Uni Mannheim/Hosp. Pegnitz, Hosp. Kulmbach
Radiodiagnosis
Uni Freiburg/Univ.-Hosp. Freiburg
Radiotherapy
Uni Würzburg/Univ.-Hosp. Würzburg
N.N.
(…/…)
N.N.
(…/…)
Scenario: Health care
© 2004 IPD, Prof. Lockemann
6
SOFSEM04
Scenario: Digital libraries
The UniCats environment
providercustomer wrapper
HTTP
trader
provider registration
user agent
HTTP
query registration and result presentation
queries to traders and trader recommendations
HTTP
HTTP
queries to wrappers and query results
order and delivery
© 2004 IPD, Prof. Lockemann
7
SOFSEM04
Scenario: Traffic prognosis
© 2004 IPD, Prof. Lockemann
8
SOFSEM04
Scenario: Traffic prognosis
ADAC
POLIZEI
M essdaten Em pfehlungen
Reiseplan
Benutzer-Profile
Logistik-Daten G roßveransta ltung
Services
Real traffic
Model world
© 2004 IPD, Prof. Lockemann
9
SOFSEM04
The optimist
Social organizations are resilient and responsive to ever changing situations because they include enough intelligent decision makers sufficient slack to take decisions.
Then why not mirror these properties to build software systems that are flexible (resilient and responsive)?
Agents are a most natural metaphor to model complex social organizations that must be highly adaptive and rely on flexible and adaptive members. Such systems are known from AI and referred to as
multiagent systems (MAS). Hypothesis: If well done, MAS offer the adaptivity and
flexibility needed for the modeled world.
© 2004 IPD, Prof. Lockemann
10
SOFSEM04
The pessimist
Multiagent systems are distributed
include asynchronous communication
have components with indeterministic behavior.
Therefore, they are inherently difficult to engineer: It is hard to give a sufficiently precise system specification
and even if there is one, what are the chances to verify or validate that the system implementation satisfies the specification?
© 2004 IPD, Prof. Lockemann
11
SOFSEM04
Where the two may meet
Are there situations with compelling reasons for a multiagent approach?
Can one yet impose constraints to be able to engineer multiagent systems?
© 2004 IPD, Prof. Lockemann
12
SOFSEM04
Agenda
Are there situations with compelling reasons for a multiagent approach? What is an agent after all? ... and a multiagent system? Useful multiagent software systems: A hypothesis Testing the hypothesis Mixed results and a refined hypothesis
Can one yet impose constraints to engineer MAS? A database person’s view At least it should be reliable! Transactional agents Reliable agent cooperation
Conclusions and challenges
© 2004 IPD, Prof. Lockemann
13
SOFSEM04
Flexibility
What is an agent?
Wooldrige and Jennings on agents:
“An agent is a computer system that is situated in some environment, and that is capable of autonomous action in its environment in order to meet its design objectives.“
Wooldrige on intelligent agents:
“An intelligent agent is a reactive, proactive, interacting agent.“
© 2004 IPD, Prof. Lockemann
14
SOFSEM04
An open system: must have an observable
effect
Effective indeterminism in time
and kind
What’s behind the definition
“An agent is a computer system that is situated in some environment, and that is capable of autonomous action in its environment in order to meet its design objectives.“
“An intelligent agent is a reactive, proactive, interacting agent.“
Agents can be many things to many people
A piece of code: software agent
Stimulators and responders
© 2004 IPD, Prof. Lockemann
15
SOFSEM04
Multiagent systems
autonomoussituatedreactive
proactive
interacting interacting
interacting
autonomoussituatedreactive
proactive
autonomoussituatedreactive
proactive
Individual objectives
Common
objective
© 2004 IPD, Prof. Lockemann
16
SOFSEM04
Multiagent systems
autonomoussituatedreactive
proactive
interacting interacting
interacting
autonomoussituatedreactive
proactive
autonomoussituatedreactive
proactive
Individual objectives
Common
objective
Can the systems be engineered?
How much flexibility can we manage?
Specialize!
Specialize!
Specialize!
Specialize!
© 2004 IPD, Prof. Lockemann
17
SOFSEM04
the programmer’s view
the software architect’s
view
the database person’s view
Who needs all that flexibility?
Hypothesis:
Multiagent systems are useful for an environment where
the range of situations (the problem space) is too large for enumeration and instead needs a qualitative description,
the problem space can be divided into sets of simpler tasks, each requiring specialized competence,
the simpler tasks can be dealt with autonomously by individual agents,
solving the overall situation requires cooperation among the agents.
the AI person’s view
© 2004 IPD, Prof. Lockemann
18
SOFSEM04
Flexibility through multiagent systems
autonomoussituatedreactive
proactive
interacting interacting
interacting
autonomoussituatedreactive
proactive
autonomoussituatedreactive
proactiveDescrip
tive
competence
Descriptiv
e
competence
Descriptiv
e
competence
Assignment of individual
responsibilities &
cooperative problem solving
© 2004 IPD, Prof. Lockemann
19
SOFSEM04
Testing the hypothesis: Scenario
MoldMaintenance
Product Assembly1
Product Assembly2
Product Assembly3
SupplierBuffer1
Buffer
MoldBuffer
A1
A2
A3
A4
A5
B1
B2
B3
B4
B5
C1
C2
C3
C4
C5
Shipping
Molding AreaProduct Assembly Area Buffer
UL01
UL02
ImmidiateBuffer2
UG01Immidiate
Buffer
UG02
UH01SupplierBuffer2 UC02
UC01
UB02
UB01UA01
UA02
UF02
UF01
Unit Assembly Area
Option
© 2004 IPD, Prof. Lockemann
20
SOFSEM04
Benchmark
Mo
ldin
g a
rea
Bu
fferProduct assembly
A
B
C
Unit assembly
Ma
nu
ala
sse
mb
ly
External supplier
SupplyOrder
Kanban
Flexibility via Kanban production organization: Production control: buffer minimization via pull principle
High variability within homogeneous product spectrum
© 2004 IPD, Prof. Lockemann
21
SOFSEM04
Benchmark
Mo
ldin
g a
rea
Bu
fferProduct assembly
A
B
C
Unit assembly
Ma
nu
ala
sse
mb
ly
External supplier
SupplyOrder
Kanban
Analytical and simulation-based layout planning for a given production program
Externally determined production schedule
© 2004 IPD, Prof. Lockemann
22
SOFSEM04
Test and Benchmark
Centrally planned Kanban is not flexible enough to deal with machine failures
short-term deviations from the delivery schedule
Solution options Passive: Simulate disturbances and adjust layout planning
Reactive: Rescheduling algorithms (known to be suboptimal)
Reactive: Agents
Comparison by simulation
© 2004 IPD, Prof. Lockemann
23
SOFSEM04
The best performing agent model
Mixed-model Assembly Line Balancing Problem (MALBP):
Reactive MAS Approach
Order Data
Order Agent Machine Agent
4: Task Announcement Bidding
6: Bid Processing
3: Task Announcement
2: Request
5: Bid
1: New Order
7: Reporting Results
© 2004 IPD, Prof. Lockemann
24
SOFSEM04
Experiments by simulation
Parameter variations: Master data:
Bill of materials Operation list fore each product
Transport lot size Affects the number of suborders for a single order
Production order generated by production planning and control
Disturbance profile per machine (statistical) Interval between disruptions Disruption duration
Output: Throughput:
Average Standard deviation
© 2004 IPD, Prof. Lockemann
25
SOFSEM04
Experiments by simulation
Throughput and Processing TimeCorresponding Standard Deviations
Parameterisation
© 2004 IPD, Prof. Lockemann
26
SOFSEM04
Experiments: Sample Analysis
Throughput Time
Standard Deviation
Suitability of MAS Compared to Centralized OR Approaches Depending on Decision Variables
Reliability and Predictability of the Results
5 min 30 min 60 min1
3
5
Disruption Duration
Lot Size
Throughput Time
1-1,5
0,5-1
0-0,5
5 min 30 min 60 min1
3
5
Disruption Duration
Lot Size
Standard Deviation
1-1,5
0,5-1
0-0,5
-1-0
Improvement by MAS
© 2004 IPD, Prof. Lockemann
27
SOFSEM04
Empirical results
Multiagent systems need slack: If assembly lines run close to capacity, MAS are even
inferior. They performed better if slack was provided by additional
machines or by delaying assembly orders.
Multiagent systems need inhomogeneity: More assembly lines by themselves have no effect unless the
machines follow very different disturbance profiles. Different disturbance profiles have even a positive effect if no
machines are added.
Multiagent systems operate best in a dynamic, non-deterministic environment: MAS are superior when it comes to evening out fluctuations
due to disruptions.
This cannot solely be explained by the original hypothesis
© 2004 IPD, Prof. Lockemann
28
SOFSEM04
Revisiting the hypothesis
Hypothesis:
Multiagent systems are useful for an environment where the range of situations (the problem space) is too large for
enumeration and instead needs a qualitative description, the range of decisions (the solution space) is
commensurate in size with the problem space, the problem space can be divided into sets of simpler
tasks, each requiring specialized competence, the simpler tasks can be dealt with autonomously by
individual agents, solving the overall situation requires cooperation among
the agents.
© 2004 IPD, Prof. Lockemann
29
SOFSEM04
Practical implementation
DB
eMPlant Simulation Model
Communication Platform using Event Mechanisms
FIPA-OS Implementation
© 2004 IPD, Prof. Lockemann
30
SOFSEM04
A database persons’ view
Hypothesis:
Multiagent systems are useful for an environment where the range of situations (the problem space) is too large for
enumeration and instead needs a qualitative description, the range of decisions (the solution space) is
commensurate in size with the problem space, the problem space can be divided into sets of simpler
tasks, each requiring specialized competence, the simpler tasks can be dealt with autonomously by
individual agents, solving the overall situation requires cooperation among
the agents.
Take a descriptive approach!After all, SQL does it!State your objectives, let the system determine the best way how to achieve them!
We need software engineering for MAS.Disciplines that could help:
Model-driven programming Service-oriented programming Component-orientation
Agent platforms are available.
But then, agents and MAS should not be indeterministic beyond their own control, such as own disturbances!
© 2004 IPD, Prof. Lockemann
31
SOFSEM04
Mastering the complexity
autonomoussituatedreactive
proactive
interacting interacting
interacting
autonomoussituatedreactive
proactive
autonomoussituatedreactive
proactive
No un-controlled disturb-ances!
© 2004 IPD, Prof. Lockemann
32
SOFSEM04
Transactions
autonomoussituatedreactive
proactive
interacting interacting
interacting
autonomoussituatedreactive
proactive
autonomoussituatedreactive
proactive
Transactional agents
Transactional conversationsAgent may be involved in concurrent conversations!
© 2004 IPD, Prof. Lockemann
33
SOFSEM04
INTERRAP BDI agent architecture
World Model
Mental Model
Cooperation Model SG PS
SG PS
SG PS
World Interface
Agent knowledge base
Cooperative planning layer
Local planning layer
Behavior based layer
SG: Situation recognition and Goal activation PS: Planning, Scheduling and execution
information accesscontrol flow
© 2004 IPD, Prof. Lockemann
34
SOFSEM04
Corresponding transaction model
action node
control node
synchronization node
parallel execution
sequential execution
alternate execution
Open nested transactions
Recovery
Conversation
© 2004 IPD, Prof. Lockemann
35
SOFSEM04
Evolutionary transactions
World Model
Mental Model
Cooperation Model SG PS
SG PS
SG PS
World Interface
Agent knowledge base
Cooperative planning layer
Local planning layer
Behavior based layer
SG
...
PS
PS
© 2004 IPD, Prof. Lockemann
36
SOFSEM04
Agent cooperation by synchronization
agent 1 transaction
agent 2 transaction
© 2004 IPD, Prof. Lockemann
37
SOFSEM04
Independent start
Agent cooperation by synchronization
M1
M12M11 M13
M111 M121... M131
...
M1111...
M112
S1
S12
S122 S123S121
S11
S0
S2
Master
Slave
Task Delegation
M11 wakes up S11
M112 waits for S1 to regain control
prevents termination of slave before
termination of master
compensate in slave if M112 fails
© 2004 IPD, Prof. Lockemann
38
SOFSEM04
Experiments with an orthogonal architecture
domain agent
local actions
database
domain agent
local actions
transactionalagent
action execution
local TA tree
common goal
distributed plan
cooperation
coordination
action(DB transaction)
robustnessservice
transactionalagent
action execution
local TA tree
database
Isolation
© 2004 IPD, Prof. Lockemann
39
SOFSEM04
Not a nice solution because …
agent 1 transaction
agent 2 transaction
Need for a shared database for conflict resolution
Cooperation protocol directly built into individual agent transaction
Asynchronous communication via ECA rules that must be coordinated with database, e.g. via active database
© 2004 IPD, Prof. Lockemann
40
SOFSEM04
distributed transaction
Transactional conversations
agent 1 transaction
agent 2 transaction
conversation
© 2004 IPD, Prof. Lockemann
41
SOFSEM04
Modeling of conversations
readyinitiator participant cfpcfp_sentcfp
refuse
not-understood
propose
prop_refused
not_understood
prop_sent
abort
deal_reached
reject-proposal
accept-proposalprop_
acceptd
prop_refused
failure
inform-done
inform-ref
initiator
participant
© 2004 IPD, Prof. Lockemann
42
SOFSEM04
Transactional conversations
TA TaskTA Task TA TaskTA Task
Transaction manager (CORBA OTS)Transaction manager (CORBA OTS)
MAS Platform IMAS Platform I MAS Platform IIMAS Platform II
User Task
DB IDB I DB IIDB II
Agent AAgent A Agent BAgent B
User Task
FIPA conformant conversation
Drawback: Conversations must be attached to the
leaf nodes
© 2004 IPD, Prof. Lockemann
43
SOFSEM04
Message-oriented middleware (MOM) with messages based on
speech acts
The next idea
agent 1 transaction
agent 2 transaction
conversation
© 2004 IPD, Prof. Lockemann
44
SOFSEM04
Conclusions: Flexibility
MAS are very costly to engineer: System-level behavior is close to impossible to predict
analytically, requires large simulative or experimental effort. The added cost does not even pay off in the majority of
cases!
The expenses seem only justified for applications with large problem and solution spaces, where the problem space is or appears non-deterministic.
A natural application seems health care. Candidate technical applications are those with a high level
of disruptions.
© 2004 IPD, Prof. Lockemann
45
SOFSEM04
Conclusions: Flexibility
Approach: Transactional properties Requires an expensive and extensive infrastructure of
database manager and transaction manager. Heavy-weight nodes needed for agents. Distributed transactions (a foe of autonomy!) and active
database mechanisms (e.g., trigggers) needed.
© 2004 IPD, Prof. Lockemann
46
SOFSEM04
Challenges: Industrial strength
Transactional synchronization: Evolutionary transactions to deal with reaction. Transactions with built-in cooperation. Non-orthogonal individual and cooperative behavior.
Transactional conversations: Difficult to attach to agent transactions. Rigid, must have ACID properties.
Message-oriented middleware: Still an unknown area for agents. Long term: A combination of transactional agents and
message-oriented middleware with higher-level guarantees. Contribute to standards!
© 2004 IPD, Prof. Lockemann
47
SOFSEM04
Conversation Policy XML (cpXML)
• Origin:
• IBM (Conversation Support for Webservices)
• Status:
• V 1.0 (August 2002)
• Specification: XML Schema, Standards document, papers, tutorial
• Coverage:
• Sequencing and timing of messages
• Tools
• Reference implementation for WebSphere
• JCA extension by Conversation Manager and Conversation Adapter
• Remarks
• Compatible to BPEL4WS
• CPs exchangeable
Roles
ConversationPolicy
SendMessageTransition
TimeoutTransition
State InitialState
LoadChild
ChildReturnTransition
Role
© 2004 IPD, Prof. Lockemann
48
SOFSEM04
FIPA-Standard
Agent Communication•Ontology Service•ACL Message Structure
Interaction Protocols•IP Library Spec
•Request, Request-When•Query•Subscribe•Propose•Contract Net, Iterated ~•English, Dutch Auction•Brokering, Recruiting
Communicative Acts•CA Library Spec
Content Languages•Content Lang. Spec•SL, CCL, KIF, RDF
Agent Management•Agent Mgmt Spec
Agent Message Transport•Msg Transport Service•Messaging Interoperability
ACL Representations•ACL Msg Representation in
•Bit Efficient•String•XML
Envelope Representations•Agent Msg Transport Envelope Representation in
•XML•Bit Efficient
Transport Protocols•Agent Msg Transport Protocol for
•IIOP•WAP•HTTP
Applications•Nomadic Application Support•Agent Software Integration
•Message Buffering•QoS•Network Mgmt
•Personal Assistant, Personal Travel Assistant•Audio-visual Entertainment and Broadcasting
Abstract Architecture•Abstract Architecture•Domains and Policies
Foundation of
Interoperable Agents
© 2004 IPD, Prof. Lockemann
49
SOFSEM04
Augmenting FIPA
FIPA-OS
Agent Management
System„White Pages“
Robustness service
Standard system DWStandard system
Scheduling
Task Task
Data Warehouse
SoftwareTask
Standard system ...
Directory Facilitator
„Yellow Pages“ Task
Data Warehouse„Agent“
Order agent Machine agent
...