Aglets Software Development Kit IBM Japo

1
Agent Workshop
Helder Coelho and Luis Moniz FCUL
2
Contents

• 1. Introduction
• 2. Workshop Design
• 3. Pilot Exercises
• 4. Conclusions

3

• 1. INTRODUCTION
• Agent
• Workshop
• Background Knowledge
• Specifications
• Standards

4
Idea of Agent

• An agent is a program that
• works autonomously during a period of time,
• interacts with the other agents and services,
• lives in an environment, and
• performs a task in name of a person or an
  organization.
• Features
• Dynamic, Adaptable, Robust and Mobile.

5
Idea of Workshop

• Space for light agent engineering, where it is
  also possible to develop innovative ideas and
  concepts along pilot exercises
• Toolboxes,
• Workbenches,
• Simulation platforms,
• Testbeds and
• Technologies.

6
Tools/Toolbox

• Specifying (AUML),
• Editing,
• Programming,
• Generating,
• Assembling,
• Monitoring, and
• Probing.

7
Workbench

• User interface.
• Agent and environment development, monitoring and
  probing tools.
• Platform capabilities.

8
Platform

• It supports the execution of agents.
• Agent management.
• Environment control.
• It supports communication
• Communication methods.
• Base structures.

9
Classification of Platforms

• According to agent types
• Isolated agents
• Simulation / Evolution
• Artificial Life
• Fixed agents
• Connection with other environments
• Local processing
• Mobile agents
• Distributed processing

10
General Specifications

• Specifications
• FIPA (Foundation for Intelligent Physical Agents)
• MASIF/OMG (Object Management Group)
• OAA (Open Agent Architecture)
• Communication languages (ACL)
• KQML (Knowledge Query and Manipulation Language)
• KIF (Knowledge Interchange Format)
• Ontolingua
• FIPA ACL
• ICL OAA

11
KQML (Knowledge Query Manipulation Language)

• Protocol for carrying and formating messages
  based in speech acts.
• Define a set of primitive messages
• tell subscribe evaluate ...
• Define the type of answer for each message
• It integrates the concept of facilitator (guide
  messages).
• The content of the messages is not defined.

12
KIF (Knowledge Interchange Format)

• Language for knowledge representation shared
  among several agents. A sort of lingua franca.
• Analogue to LISP declarative
• Expressiveness of first order logic
• (salary john 550)
• (gt ( (length farm1) (width farm1)) ( (length
  farm2) (width farm2)))
• (gt (and (real ?x) (integer ?n)) (gt (exp ?x ?n)
  0))

13
Ontology

• Aim represent the knowledge structure
• Easy to use, ability to process knowledge.
• Semantic web
• Varieties
• XML/XML Schema
• RDF/RDF Schema
• OML
• DAML, DAMLOIL, DAML-S, DAML-L
• ...

14
Foundation for Intelligent Physical Agents (FIPA)

• FIPA International Association of enterprises to
  share general specifications on agent technology.
• Standards for
• Concepts
• Architectures
• Services
• Communication
• Platforms
• ....

15
FIPA Platforms
(www.fipa.org/resources/livesystems.html)

• Specification for implementing platforms
• Obligatory request
• AMS (Agent Management System)
• DF (Directory Facilitator)
• ACC (Agent Communication Channel)
• Support of
• Link between platforms via IIOP protocol
• Distribution of a platform among several machines
• ....

16
FIPA Platform
17
OMG (Object Management Group)

• MASIF (Mobile Agent System Interoperability
  Facility)
• Common platform for allowing communication among
  mobile agents.

18
CORBA/ORB
Integration of agents in ORB architectures
19

• 2. Workshop Design
• Platforms
• Workbenches
• Testbeds
• Technologies
• Commercial Products

20
Simulation Platforms

• Swarm
• Tileworld
• Javabots
• Teambots
• Mice
• ABE
• MACE
• TouringMachines
• ViewGen

• Tierra
• StarLogo
• EINSTein
• RoboSoccer
• RoboRescue
• SimCog
• Creatures
• SIMS
• ........

21
Workbenches

• Agent Tcl (Dartmouth)
• April (Fujitsu)
• Clearlake (Guideware)
• M0 (Univ. of Geneva)
• Obliq (DEC SRC)
• Safe-Tcl (N. Borenstein)
• Tacoma (Univ. of Norway Cornell Univ.)
• Telescript (General Magic)
• Wave (Univ. of Surrey)

• Internet System Environment (OSF)
• Orbix (Iona)
• Sumatra (UMCP)
• JAM (Univ Columbia)
• JIAC (DAI-Lab/Tech.Univ. Berlin)
• EXCALIBUR (Tech. Univ. Berlin)
• FarGo (Israel Inst. of Tech.)
• Ara (Univ. Kaiserslautern)
• Agent CLIPS (YilSoft)

22
Workbenches

• Aglets (IBM TRL)
• Concordia (Mitsubishi)
• ABE (IBM)
• CyberAgents (FTP Software)
• Odyssey (General Magic)
• MOA (MASIF/OMG)
• Voyager (ObjectSpace)
• ZEUS (BT)
• Hive (MIT)
• Bee-Gent (TOSHIBA)

• Agent Building Shell (Univ of Toronto )
• A-Match (CMU)
• RETSINA (CMU)
• DECAF (Univ Delaware)
• Mole (Univ. Stuttgart)
• OrbixWeb (Iona)
• Java Agent Template (Stanford)
• OAA (SRI)
• .....

23
Workbenches (FIPA)

• Agent Development Kit
• April Agent Platform (Fujitsu)
• Comtec Agent Platform
• FIPA-OS (Nortel)
• Grasshopper
• JACK Intelligent Agents
• JADE (TILAB)

• JAS API (Fujitsu, Sun, IBM, HP, ...)
• LEAP
• ZEUS (British Telecom)
• XDepint
• SACI
• NETDEF

24
Testbeds

• Exercices
• SEM
• Depint
• SEM
• Languages
• Prolog, StarLogo, CLIPS
• C, Java

25
Technologies

• Symbolic Programming
• Genetic Programming
• Ant Programming
• Neural Networks
• Fuzzy Logic
• Cellular Automata

26
Look to Isolated Agents

• Reduced environments.
• Simple agents.
• A big number of simultaneous agents.
• No external link.
• Artificial Life, Simulation, Evolution and
  Learning.
• Examples
• Swarm, StarLogo, TileWorld, EINSTein, Javabots,
  RoboSoccer.
• Creatures, SIMS, ....

27
StarLogo

• Based on LOGO.
• Fast prototyping of applications.
• Several simultaneous agents.
• Identical agents (same code).
• Very simple agents
• Emergency of global behaviours upon isolated ones.

28
EINSTein

• Environment for simulation military battles.
• Agents defined upon fixed parameters
• Behaviour
• Personalidade
• Comunicação
• Two groups facing each other.
• Territory specification.

29
RoboSoccer

• Client/Server.
• SoccerServer
• It executs the players commands.
• It sends for each player environment information.
• It controls the environment integrity.
• Clients (Players).
• They process the received information.
• They send commands to the server.

30
MABS Platforms
31
Look to Fixed Agents

• Heterogeneous agents.
• Server.
• Support for communication.
• Distributed environment.
• Examples
• AgentCLIPS, JAT.

32
Agent CLIPS

• Agents written in CLIPS.
• Link to NNTP and HTTP servers.
• Common use
• Access to web/news pages
• Data filtering
• Communication with other CLIPS agents.
• Available MacIntosh and Unix (Sun).

33
Java Agent Template (JATLite)

• For building agents that talk through the
  Internet.
• Non migratory (yet achieved with Java RMI).
• Communication with KQML.
• Open architecture, designed for allowing easy
  extensions with new funcionalities.

34
Open Agent Architecture (SRI)

• Agent facilitator
• Blackboard
• Register service
• Message chaining
• Agent linking
• Communication language

35
Open Agent Architecture (OAA)

• Agents written in PROLOG/JAVA.
• When registering, the agent shows the services
  for the community.
• When receiving a request, the facilitator
  forwards it to the agents able to solve it.
• Requests with querys PROLOG form.
• No control over the agents.

36
Open Agent Architecture (OAA)
37
Open Agent Architecture (OAA)
Facilitator Agent
Registry
Interagent Communication Language
Modality Agents
38
JADE (Java Agent DEvelopment Framework)

• It supports FIPA standard for agent platforms
  AMSDFACC.
• To each machine is associated an agent container
  (a sort of JVM).
• A platform may have several containers
  communicating via RMI.
• Agents written in JAVA.
• A GUI controls several agents
• Start, suspend, kill ....

39
JADE (Java Agent DEvelopment Framework)
40
Look to Mobile Agents

• Environments for supporting distributed code.
• Agent movement among machines
• Reduction of net traffic
• Elimination of comunication delays
• Autonomous and syncronous execution
• Dynamic adaptation to problems
• Heterogeneous environments
• .....
• Examples
• Aglets, Voyager, Odissey, DAgents, Ara.

41
Safe TCL

• Standard for the creation of agents.
• Agents are defined using a subset of Tcl
  language, allowing the server security.
• Transport means email.
• Support system Tcl/tk Safe Tcl MIME.
• Simple system to create and send agents to remote
  locations.

42
Telescript

• First commercial language oriented to create
  mobile agents.
• The agents correspond to Telescript objects and
  they conserve their state when they move.
• Some disadvantages
• Require several system resources
• Commercial (costs)
• Out of market General Magic is developping
  agents in Java.

43
DAgents (Agent Tcl)
http//www.cs.dartmouth.edu/agent/

• Plataform for mobile agents
• An agent is a script in Tcl
• It allows the movement of the agent using
  agent_jump, transporting the agent plus its
  execution state
• Security is achieved with agent encryptation and
  Safe-Tcl
• Extensions Scheme, Java and Phyton.
• Available Unix.

44
Agent Building Environment (ABE)

• Development workbench.
• Facilities for building up applications based in
  distributed/mobile environment.
• Adopted for services
• NTTP, HTTP, Time, File
• New adaptators in Java or C.
• Available OS2, Windows, AIX.

45
Aglets Software Development Kit (IBM
Japan/OpenSource)
http//www.trl.ibm.co.jp/aglets

• Integrated package for design and management of
  multi-agent systems.
• New communication protocol ATP (Agent Transfer
  Protocol).
• Agent server (Tahiti) creation, remotion ....
• Language JAVA.
• Extensions JMT (Java-based Moderator Template),
  JKQML (Java KQML).
• Available SPARC/Solaris 2.5, x86/Solaris(1.1
  beta 4), Windows95/NT, AIX 4.1.4, OS/2(1.1.1
  beta).

46
Aglets Software Development Kit

• Model fixed execution based in events.
• Creation, Remotion, Mobility, Persistence.
• onCreation(), run(), handleMessage(),
  onArrival(), onDeparting(), onActivation() .....
• Agents executed within the platform (Tahiti).
• Several JAVA restrictions
• Control of resources.
• Recover of execution states.

47
Aglets Software Development Kit

• Example

The return of WriterSlave depends
on Timeout User interaction Writer informs what
is the result of the operation.
48
Ara
http//www.uni-kl.de/AG-Nehmer/Projekte/Ara/index_
e.html

• Platform for safe execution of agents in
  heterogeneous nets.
• Language TCL or C/C (pre-processed).
• Available SunOS, Solaris and Linux.

49
Mole
http//www.informatik.uni-stuttgart.de/ipvr/vs/pro
jekte/mole.html

• It supports agents written in Java.
• Functionalities
• location - basic services for agents.
• engine - services for locations.
• Types of agents
• System agents
• User agents
• Common agents

50
Odyssey(General Magic)
http//www.genmagic.com

• System of agents implemented as a set of JAVA
  libraries.
• It supports the concepts of agent and location.
• It does not support the transport of the agent
  state.
• Platform JDK1.1

51
Voyager(ObjectSpace)
http//www.objectspace.com

• It is an ORB written in Java.
• Oriented for building distributed applications.
• It supports the CORBA standard.
• Agents are treated as Objects.
• Integration with applets.
• It is not a platform specific for agents.

52
Voyager(ObjectSpace)

• Example

• Each shop has the same set of products.
• Agent SHOP searches the shop that has the
  desired and cheapest product and stops there.
• Agent BUY search the shop where SHOP is stopped
  and buys the product.
• Agente SHOP goes away to search again for the
  best price.

53
NETDEF

• Parallel high level language.
• Compiler able to translate a NETDEF program into
  a neural network.
• Computation equivalence between a neural net and
  a classical program.
• Symbolic and subsymbolic computations are closer.
• Main features modularity, syncronism and massive
  parallelism.
• Web site www.di.fc.ul.pt/jpn/netdef/netdef.htm

54
Look to Commercial Agents

• Several tentatives to produce commercial products
  based on agent tools
• Archon
• Telescript, Odissey, Serengueti, Portico
• Creatures, The Sims
• Examples OrbixWeb, Serengeti and Portico.

55
OrbixWeb(Iona Technologies)

• Platform for developping client/server
  applications according the CORBA standard.
• It generates Java applications.
• It is not specific for agent construction.
• Commercial product.

56
Serengeti (General Magic/MicroSoft)

• Personal assistant.
• VUI/GUI user interface.
• To be partially integrated with the next
  Microsoft Outlook version.
• Technology
• Odyssey (?).
• Actual state (?).

57
Portico (General Magic)

• Portico is a virtual assistant and an integration
  and access factor for several data and
  communication services.
• Functionalities
• Speech recognition
• Conversation handling
• Data filtering
• Agenda management
• Data search
• Technology
• Odyssey
• MagicTalk

58

• 3. Pilot Exercises
• GIA and LabMAG
• (Helder Coelho, Leonel Nóbrega, Luis Moniz, Paulo
  Urbano).

59
Exercise1 Training soccer tactics

• Applying the ideas of the society of mental
  states framework (Corrêa and Coelho, 1998).
• Generating special soccer players with the
  RoboSoccer platform need of new tools.
• Tool package Agent editor, Agent code generator,
  Agent executor and Agent behaviour probing.
• Experiment to coach the movements and actions of
  a partcular soccer team (Nóbrega, 2001).

60
Exercise 2 Tailoring teams

• Extension of the OAA platform with 3 components
  for automatic agent generation Agent listener,
  Agent creator and Agent activator.
• Agents regarded as service providers with a
  public ontology interface, externally shared by
  the platform.
• Focus on complex, dynamic and adaptative group
  behaviours and tasks agents as building blocks
  of intelligent and customised applications (Moniz
  and Coelho, 2001).

61
Exercise 2 Tailoring teams

• Architecture of the 3 system structure levels
  Service (implementation and management of the
  agent and service black box), Design
  (construction of applications by aggregating the
  resources as a community of agents in the Service
  level) and User (gateway to the system
  manipulation in terms of task specifications and
  high-level services).
• Experiment find a movie by decomposing this
  goal as a set of tasks to be solved by a group of
  agents.

62
Exercise 3 Mixing social and physical
features

• Educational environment for behaviour based
  robots focus on advanced experiments on
  collective behaviour.
• Testebed mixing Aglets and Player/Stage.
• Extended Aglets multi-agent platform to arrange
  better tools for the designer.
• Player/Stage environment for robotics simulation.
• AGLIPS linkage between those two heterogeneous
  environments (Aglets and Player/Stage) to build a
  virtual laboratory robots are specified in CLIPS.

63
Exercise 3 Mixing social and physical
features

• Composite behaviour according to the idea of
  Brookss subsumption architecture (Moniz et al,
  2002).
• Experiment arrange a library of agents, control
  programs and environments (number, size, location
  of obstacles, number and location of robots, and
  behaviour of each robot) to build flocking
  behaviour in a world full of obstacles.
• Web site ai.di.fc.ul.pt/projects

64
Exercise 4 Brokering service agencies

• Extended Aglets platform for delivering new
  services (white pages, group formation and yellow
  pages) in a distributed and mobile agent
  environment.
• Extension along 4 lines environment,
  architecture, live cycle model and communication
  (message model to support KQML like directives
  and speech acts).
• Complex robot communication channel with white
  pages service and GPS like facility (Moniz et al,
  2002)..

65

• 4. Conclusions

66
Trends

• Tools
• Agent construction/design
• Agent control/management
• Agent monitoring/probing
• Associated with an objective/task.
• Integration and distribution.
• The majority not yet commercial.

67
Today

• Recent technology
• JINI
• BLUETOOTH
• UTMS
• Current Commercial Applications
• Animation (Antz, Shrek, ...)
• Games (Age of Empires, Creatures, Magestic, Sims)
• Personal assistants (PDAs)
• Automatic net management
• Data and text retrieval
• Economy and management towards complexity!