AO Software Engineering:

1
AO Software Engineering

• The Challenge of Realworld Business Scenarios,
  and Applications

2
Overview

• realworld business applications the challenges
• AO software development
• success stories (?)

New German Research
Program "Intelligent Agents Realworld Business
Scenarios"
3
Realworld business applications the challenges

• flexibility
• the "right" problem
• already existing solutions
• towards problem-centered human-computer
  cooperation
• standards
• scalability

4
Flexibility

• be faster!
• online-adaptation of business processes
• coordination from hierarchies to decentralized,
  local control
• locally autonomous units, and "fractalization" of
  enterprises
• re-inventing the value chains
• virtual enterprises

5
The "right" problem ?

• modeling the world vs. implementing a
  software solution
• methods (planning, scheduling, etc.) vs.
  domains (banking, manufacturing, etc.)
• large applications vs. small add-ons to
  existing systems
• closed local network without any external
  access vs. fully open, possibly world-wide
  network

6
Already existing solutions

• meet the benchmarks! ... ??
• competitive advantage of AOIS?
• how to incorporate agents into large business
  applications?
• security
• examples
• - workflow management
• - planning, scheduling, simulation
• - telecommunication

7
Problem-centered human-computer cooperation

• HCCW (Steiner et al. 1989), but
• interface technology for agents?
• successful ontologies?
• "hybrid" cooperative problem solving / inference
  processes?
• communication languages adequate for humans?
• organizational perspective?
• social, and sociological perspective?

8
Standards

• Knowledge Query Manipulation Language (KQML),
  1991
• Knowledge Interchange Format (KIF), 1991
• Foundation for Intelligent Physical Agents
  (FIPA), 1997
• Mobile Agent System Interoperability Facility
  Specification (MASIF), 1997
• ? Does agent technology also meet the industrial
  standards (software technology standards, domain
  standards)?

9
Scalability

• how many agents ...
• interacting in a typical research lab
  application?
• to be managed on an agent platform?
• necessary for a particular application?
• being expected in an open environment (internet)?
• and already under control today ??

10
Status

• Agents ...
• provide an exciting field of research!
• push new concepts without any own substantial
  commercial success ...?
• may sometimes require external push (e.g., Java,
  mobile agents, CORBA, internet/WWW, XML,
  e-commerce) ...?
• Is the technology able to push by itself its own
  success??

11
AO software development

• moduls gt objects gt components gt agents ?
• AO software development methodologies
• architectures agents / multiagent systems
• modeling and implementation languages
• testing, maintenance
• life cycle models for agents / multiagent systems

12
Extensions of object-oriented methodologies

• AO analysis design (Burmeister 96)
• agent modelling for BDI agents (Kinny, Rao,
  Georgeff 96)
• MA Scenario-Based Method (Moulin 94, 96)
• AO methodology for enterprise modelling (Kendall
  96)

• Problem agents are different from objects!
• interaction not just through msg passing/method
  invocation
• agents have mental states, inferences, (deep)
  knowledge, etc.
• social dimension

13
Extensions of knowledge-engineering methodologies

• CoCoMAS (Glaser 1996)
• MAS-CommonKADS (Iglesias 1997)

• Problems
• methods for knowledge acquisition, modelling and
  reuse
• knowledge distribution, communication and
  interpretation
• social dimension of MA systems

14
Agent architectures
15
Reactive Agents
World
Controller
Stimuli
Plans
E f f e c t o r
S e n s o r
. . .
. . .
Agent
16
Deliberative Agents
World
Cognition
Inference Strategies
Memory
E f f e c t o r
S e n s o r
Utility Function
Inter- pretation
Agent
17
Interacting Agents
World
Coordination
Agent
Agent
Interaction
Interaction
Interaction
Interaction
Interaction
Interaction
Interaction
Interaction
Agent
Agent
18
Hybrid Agents
Cognition
World
Inference Strategies
Memory
Goals
Environment Model
Utility Function
Inter- pretation
Domain Knowledge
Planner
Stimuli
E f f e c t o r
S e n s o r
Agent
Pattern 1
Plans
Plan 1
Pattern 2
. . .
. . .
Plan n
Pattern n
19
Status

• "toolbox" of agent architecture types available
• benchmarking of agent architectures?
• agent architecture design as an engineering
  discipline?
• (proven) standards for agent architectures?
• which architecture for which problem?
• agent architectures vs. related "non-agent"
  architectures (client/server, CORBA, etc.)?
• agent architectures vs. MA systems
  architectures?

20
Success stories ?
21
Some examples

• Network Management at German Telecom
  hierarchical coordination of network traffic
• mobile agentscan they be complex, deliberative
  agents?
• planning, scheduling simulationadvantages
  compared to currently used solutions?
• information agentscooperative problem solving?
• workflow managementcompetitive advantage
  through agent technology?

22
New German Research ProgramIntelligent Agents
and Realistic Business Scenarios

• task to be addressed distributed / multiagent
  planning of business processes in logistics
  (goods informations)
• domains manufacturing vs. health care
• approach collaboration of management science,
  information systems, computer science
• aim evaluating/adapting agent technology
  against/to industrial requirements

23
Research process
24
T1 Methodological Foundations

• fomal representation and evaluation of knowledge
  in management science
• reference modelling, e.g. CIM-OSA, ARIS, SOM
• modelling languages, e.g. petri-nets, UML

25
T2 Application Scenarios

• domains manufacturing, health care
• creation of realistic application models
• study / evaluation / comparison of typical
  application problems potentially be addressed by
  agent technology
• definition of benchmarks
• ontologie

26
T3 Agent Technology

• agent modelling
• agent architectures
• coordination languages
• communication languages
• design methods/design tools
• agent platforms

27
T4 How do agents affect business processes?

• inter-/intra-organisational value chains
• management decision making
• organisational structures
• organisational culture
• technology-driven organisational change
• innovation strategies for agent technology

28
Intelligent Agents and Realistic Business
Scenarios

• duration 7/2000 - 6/2006
• budget 18 projects
• coordinator Stefan Kirn, TU Ilmenau
• URL www.wirtschaft.tu-ilmenau.de/wi/wi2/SPP-Agent
  en/