Autonomous Agents for Flexible Hypermedia Systems

Autonomous Agents for
Flexible Hypermedia Systems
(and the other way ‘round)
Simon Mayer
6th International Workshop on Engineering Multi-Agent Systems
Stockholm, Sweden, July 2018

July 2018Page 2 Autonomous Agents for Flexible Hypermedia Systems | Simon Mayer
Interacting Smart Environments
Flexible Integration and Reconfiguration
Cooperation with People
Servitization of Resources

Interacting Smart Environments
Flexible Integration and Reconfiguration
Cooperation with People
Servitization of Resources
Intelligibility for People

Talk in a Nutshell
1. Considering patterns from the Web architecture
and in particular the Web of Things is a good idea
for engineering global MAS
- Uniform interface
- HATEOAS principle
- For free: openness, easy integration, human-friendliness
2. Combining with the WoT also yields lots of
relevant real-world use cases
3. The WoT community moved toward autonomous
agents over the past years, and will welcome
these advances

Talk Outline
1. Hypermedia-based Resource Integration
2. The Web of Things and selected aspects of
the Web Architecture
3. Building Bridges between WoT and MAS:
Orchestrating Industrial Manufacturing
4. Hypermedia-based Agent Communities
beyond Industry

Motivation: Toward a lot-size-one World
Today: Highly optimized, static, inflexible production lines (>30yr). Upgrading disrupts production!
Tomorrow: Lot-size-one dilemma – every product is unique but production still must be cheap

Toward a lot-size-one World: Challenges
Traditional factory integration: manual and expensive
− Prone to supplier lock-in: supplier of most expensive device locks customers into its ecosystem
− Inflexible links between resources (addressing + protocol conflicts, manual configuration, etc.)
Mass customization at small lot sizes requires automatic reconfiguration
− Loose coupling between modular components → evolvability → innovation!
− Links need to be changed dynamically by people or software!
Our strategy: Use principles from the World Wide Web to integrate devices
[Mayer et al.: UberManufacturing: A Goal-Driven Collaborative Industrial Manufacturing Marketplace. IOT 2016: 111-119, 2016]

The Web of Things
Original WoT idea: Applying Web principles to IoT devices makes them usable by people!
− The Web Architecture and its technologies (i.e., http & friends) are proven and scale!
− Virtually all devices (even class 1, at 100/10 kB) can run (crude) Web servers!
− There are libraries for virtually all programming languages!
− Rather simple to secure and authenticate
− Catalyzed by increasing reliance on external developers to build innovative services
− The Web is incredibly easy to use – this is key to adoption of an IT system
Perfect!
Why does no one think about me?
[F. D. Davis. Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology, 1989]
[D. Gefen and M. Keil. The Impact of Developer Responsiveness on Perceptions of Usefulness and Ease of Use, 1998]
[D. Guinard, I. Ion, S. Mayer: REST or WS-*? A Developers’ Perspective, 2011]
[L. Popa, P. Wendell, A. Ghodsi, I. Stoica: HTTP: An Evolvable Narrow Waist for the Future Internet, 2012]

The Web Architecture
From a networking perspective, IP enables heterogeneous systems to talk to each other: it’s the stack’s “narrow waist”
In the “Internet of People”, the Web is already a “narrow head” of the stack (to some extent even replacing email!)
…and it is well on track to becoming a narrow head of the networking stack in IoT scenarios as well
Web
(HTTP, CoAP)
Source: Steve Deering, Talk at IETF 51, 2001

WoT Issue #1: Things’ Idiosyncrasies
The IoT's edge devices can look very different from the Internet of Computers
- Low bandwidth and intermittent connectivity
- Transmissions tend to be more “bursty” in IoT settings
- Often, multicast is desired as one wants to talk to many sensors / actuators at once
- User perspective: “It’s a bunch of traffic sensors – I don’t care what their duty cycle is, I’ll treat them as always-on”
The Web can help: Writing IoT applications shouldn’t require knowing these details
- Abstraction is exactly what the Web is good at! :-)
- We can put IP and HTTP stacks on very small devices (100KB / 10KB)
- Web protocols run well on top of constrained stacks (e.g., via 6LoWPAN)
- Web protocols (CoAP) enable multicast and can deal with intermittent connectivity

WoT Issue #2: Interconnecting Things across IoT Silos
How can we interconnect (“mashup”) applications across walled gardens?
IoT scenarios in many cases involve broad and deep scenarios
- Broad in the sense that resources come from a variety of sources/manufacturers/providers
- Deep in the sense that they involve taking multiple steps / using multiple APIs
The Web can help: Openness is at the very core of the Web architecture!
- The Web reduces coupling between servers and clients
- New entrants can fully exploit (legacy) applications from the start at a well-defined entry point (“bookmark”)
- All required a priori knowledge is standardized (e.g., as part of REST)
- Resources and applications can evolve independently
Deliver some steak by
6pm please!
Stop! We take burger instead :-|

WoT Issue #3: Enable Machines to use IoT Functionality
In the IoT, we want to enable machines to use provided functionality
- What would the Social Web be if People couldn’t navigate it?
- What will the Internet of Things be if Devices can’t navigate it?
The Web can help!
- Shared information models (“machine understanding”) are required for doing this - how to define those is
one of the main challenges in the WoT / IoT / AAMAS / Semantic Web / AI / Applied Philosophy space!
- But at least the Web allows us to ignore low-level protocol semantics since these are already defined
within the Web’s architectural style (i.e., REST)
Imagine programming a client so that it uses Web APIs like you would…

WoT Issue #3: Enable Machines to use IoT Functionality
How does a person decide which link to follow on a page?
<a href=“somewhere”>something</a>
1. Read “something”
2. Think “something” is really interesting (given their belief state)
3. Click “something”
This isn’t trivial for machines at all…
- The machine would need a deeper understanding of the high-level semantics of the domain in question
that we acquire by years of experience in the real world
- Semantically annotating links in a machine-readable way is no general easy way out…

The Web of Things
Today’s WoT idea: The Web’s properties also make APIs simpler to use for machines!
…and everything integrates with the human Web (i.e. browsers)
…and we inherit the Web’s other proven mechanisms for free!
Wait, what? Machines that “use” other
machines? Sounds like we should
talk to those guys at AAMAS!
Sounds great!!

The REST Architectural Style
The Web’s Architectural Style is called “Representational State Transfer”
And that is not (only) a buzzacronym
The Web Architecture focuses on three central pillars
1. Shared Identification Model that allows everybody to refer to available resources (URI)
2. Self-Describing Representations that allow clients to understand resources (e.g., HTML/XML/JSON)
3. Uniform Interfaces that can be used to access those resources (HTTP & Friends)
And then there is two more aspects to REST that are a little more intricate
− Stateless Interactions
− Hypermedia Driving Application State (“HATEOAS”)
These principles enable the Web’s beneficial features scalability, mashup-ability, usability, accessibility, etc.

Web Architecture: Uniform Interface
The same (small) set of operations applies to every resource
A small set of verbs applied to a large set of nouns
− Verbs are universal and not invented on a per-application base
− New nouns are invented for each application, but are universally identified
− Verbs are polymorphic and apply to all nouns
− Why should we care?
− Because verbs enter languages much more slowly than nouns
− A language that reflects this is more robust
HTTP works in this way (by defining a controlled set of HTTP methods)!
− Other schemes have different (or no) interaction models
− ftp: works similar to HTTP (TCP/IP-based access with a defined set of commands)
− isbn: does not implement an interaction model with the identified resource

REST: Defined set of verbs with fixed semantics! (e.g., GET, PUT, DELETE, etc. for HTTP)
− Find current temperature: GET thermometer/temperature
− Load size of an image: GET image23/size
− Verbs are polymorphic in REST!
WS-*: Definition of operations that can be applied to objects
− Find current temperature: TempWebService.getTemp(…)
− Load size of an image: ImageWebService.calcImgSize(…)
− Verbs are not polymorphic in WS-*!
TempWebService
getTemp 60°F
ImageWebService
calcImgSize 200px
TempWebService
calcImgSize FAULT
[http://www.sunspotworld.com]

Uniform interfaces enable generic clients!
Every Web browser on the planet can access every Web resource on the planet!
− This is because HTTP works in the same way for every resource
− Interacting with the resource is possible without knowing it beforehand
Consciously select which operation to use for an operation on a resource
− HTTP GET and HTTP HEAD are safe (= side-effect free) operations
− HTTP PUT and HTTP DELETE are idempotent operations
− HTTP POST can have side-effects
Caching + Pre-fetching!
Safe retries!
None of this!

Web Architecture: Uniform Interface and MAS
Hypermedia-embedded MAS would inherit the uniform interface
Directly inherit performance gains (caching, pre-fetching, safe retries)
Exploit fixed low-level semantics of the Uniform Interface!
- If something is a GET, it is side-effect free
- If something is a PUT and the parameters are static, it is idempotent
i.e., a GET will never change a (core resource) state variable!
i.e., a repeated PUT will never change any (core)
state variable (or any other relevant belief state)

REST: Hypermedia as the Engine of Application State
Servers guide interactions by providing links within resource representations.
Clients use resources by navigating the provided links.
REST applications navigate instead of calling
− Representations contain information about possible traversals
− The application navigates to the next resource depending on link semantics
− Navigation can be delegated since all links use uniform identifiers
This is more important than it seems… especially for machines
− Server provides local guidance for its clients by providing links
− Those encode all typical state transitions that a server foresees for its clients
In the Web/WoT world, the central challenge of “autonomous agents” is how to build
global guidance by assembling distributed locally guided HATEOAS systems

Important part of REST! Especially for service integration…
HATEOAS

“Prominent members
include the Hypertext
Transfer Protocol.”
http://en.wikipedia.org/wiki/HTTP
http://en.wikipedia.org/wiki/Networking_protocol
http://en.wikipedia.org/wiki/WWW
http://en.wikipedia.org/wiki/URI
http://en.wikipedia.org/wiki/Hyperlink
GET Known URL!
“The Hypertext Transfer
Protocol (HTTP) is a
networking protocol for
distributed, collaborative,
hypermedia information
systems. HTTP is the
foundation of data
communication for the World
Wide Web.”

That is a (projection of a) state machine!
The transitions between states are REST
operations that are provided by resources via
hyperlinks! → This is HATEOAS
POST z

Did you ever experience that the “back”
button broke a hypermedia application?
e-banking, e-commerce, flights booking, etc.
Clients that provide a “back” button break the HATEOAS contract!
− The server didn’t want a “back” link to be available in the application
− If it had wanted one, there’d have been a hyperlink!

HATEOAS is what adds flexibility to REST systems!
HATEOAS is what adds robustness to REST systems!
− Links are discovered by clients at runtime
− Clients will automatically adapt if links change!
− This is why browsers are general-purpose tools and not specialized applications
Browsers are not compiled
against Hypermedia APIs!

http://bookstore.org/carts/simon http://bookstore.org/books/book1
http://bookstore.org/checkout?customerID=simon
Known URL!
“Recommended Books: Book1
Books in your cart: None.
Checkout not possible (no books).”
“This is Book1. It’s great!
Add book to cart”
GET books/book1
POST carts/simon?book=Book1
“Recommended Books: None
Books in your cart: Book1.
Go to checkout.”
“Select payment type:
Credit Card, Debit Card”

http://bookstore.org/carts/simon
Known URL!
http://bookstore.org/books/book1 http://anything.org/things/book1
This (tiny) change would break any client that is
statically compiled against the book store API….

http://bookstore.org/carts/simon http://anything.org/things/book1
http://bookstore.org/checkout?customerID=simon
Known URL!
“Recommended Books: Book1
Books in your cart: None.
Checkout not possible (no books).”
“This is Book1. It’s great!
Add book to cart”
GET things/book1
Host: anything.org
POST carts/simon?book=Book1
Host: bookstore.org
“Recommended Books: None
Books in your cart: Book1.
Go to checkout.”
“Select payment type:
Credit Card, Debit Card”

To guide clients in applications, the server provides hyperlinks that they may follow
− However, servers can only provide hyperlinks they know about (e.g., industrial automation)
− However, servers will only provide hyperlinks they want the client to know about (e.g., integration with competitors)
We view each HATEOAS system as an information silo
Enabling global guidance for machine agents = bridging these information silos
I can do that!
What about me??
*hope*

The Web of Things
REST enables scalability, mashup-ability, usability, accessibility
Those aspects are very important for the Web (of Documents) and the Web 2.0 (of People)
They are even more important for interacting physical devices in the IoT and in UbiComp scenarios!
− Machines cannot easily recover if something goes wrong in an interaction
− Machine clients can depend on REST to provide the low-level interaction semantics for all resources on the Web
via its Uniform Interface and HATEOAS
− To make the Web (of Things) navigable for machine clients, we “only” need to solve the problem of interoperability
on the level of high-level domain semantics
…which brings us back to our initial problem…

Back to the lot-size-one manufacturing idea…
Traditional factory integration: manual and expensive
− Prone to supplier lock-in: supplier of most expensive device locks customers into its ecosystem
− Inflexible links between resources (addressing + protocol conflicts, manual configuration, etc.)
Mass customization at small lot sizes requires automatic reconfiguration
− Loose coupling between modular components → evolvability → innovation!
− Links need to be changed dynamically by people or software!
Strategy: Use principles from the World Wide Web to integrate devices
[Mayer et al.: UberManufacturing: A Goal-Driven Collaborative Industrial Manufacturing Marketplace. IOT 2016: 111-119, 2016]

Combine hypermedia with coordination technologies
from multi-agent systems and with automatic logical
reasoning over embedded functional profiles
Note: this approach directly applies to any hypermedia system: e-commerce; e-government; online social media; etc.
Mass customization through automatic reconfiguration of production
systems based on a given production goal and non-functional parameters

Automatic Reasoning over Embedded Functional Profiles
ArcMate 100iC/12S
“I can perform arc welding”
Collaborative Robot
“I can pick-and-place objects”
Object is reachable
Object has been repositioned
HTTP PUT to
myRobot.org
𝑃𝑟𝑒𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 → 𝑃𝑜𝑠𝑡𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ∧ 𝑊𝑒𝑏 𝑅𝑒𝑞𝑢𝑒𝑠𝑡
[Mayer et al.: Smart Configuration of Smart Environments. IEEE Trans. Automation Science and Engineering 13(3), 2016]

Technical Approach
{
Precondition
}
=>
{
Postcondition
Web Request
}.
Collaborative Robot
ArcMate 100iC/12S
“I can perform arc welding” Example: Notation3 Format

Technical Approach
{
?state a st:State; log:includes {
?anObject ex:locatedIn ?sourceLocation.
}
?targetLocation a ex:Location; ex:X ‘0.55’; ex:Y ‘-0.46’; (...)
}
=>
{
Postcondition
Web Request
}.
Collaborative Robot
ArcMate 100iC/12S

Technical Approach
{
}
}
=>
{
[ a st:StateChange;
st:replaced { ?anObject ex:locatedIn ?targetLocation. };
st:parent ?state ].
Web Request
}.
Collaborative Robot
ArcMate 100iC/12S

Technical Approach
{
}
}
=>
{
[ a st:StateChange;
st:replaced { ?anObject ex:locatedIn ?targetLocation. };
st:parent ?state ].
_:request (...) // Standard description of the Web request that enacts the transition
}.
Collaborative Robot
ArcMate 100iC/12S

Collaborative Robot
ArcMate 100iC/12S
Technical Approach
Use logic planning to chain
functional descriptions of the
resources and construct a
hypermedia path from A to B
{
_:initialState a st:State; log:includes {
ex:Box1 ex:locatedIn ex:InitialLocation.
}
ex:FinalLocation
ex:id http://example.org/locations/iiwa-ee;
ex:X “0.55”; ex:Y “-0.46”; ex:Z “0.60”;
ex:OX “0”; ex:OY “0”; ex:OZ “-0.7071”; ex:OW “0.7071”.
(...information about other locations etc.)
}
A
{
_:goalState a st:State; log:includes {
ex:Box1 ex:locatedIn ex:FinalLocation.
ex:Box1 ex:foldingMethod ex:SCF.
ex:Box1 ex:sealingMethod ex:SSM.
}
}
B
Initial world state
Target world state

Automatic Reasoning: Integrating People
Maintenance Specialist
“I can repair arc welding robots”
Commissioning Expert
“I can commission collaborative robots”
[Mayer et al., An Open Semantic Framework for the Industrial Internet of Things, IEEE Intelligent Systems 32(1), 2017]
ArcMate 100iC/12S
Collaborative Robot

Integrating People
Automatic Reasoning: Human Workers
Commissioning Expert
“I can commission collaborative robots”

Integrating People
Automatic Reasoning: Human Workers and Workplace Safety
e.g., kneeling, lifting, over-head-holding, etc.
Body & Medical
Foundational Model of Anatomy
Tools, Products, Processes
ISA-88, DIN 2860, DIN 8580
Health & Safety

Integrating People
Automatic Reasoning: Human Workers and Workplace Safety
e.g., kneeling, lifting, over-head-holding, etc.
Body & Medical
Foundational Model of Anatomy
Tools, Products, Processes
ISA-88, DIN 2860, DIN 8580
Health & Safety
Remind human workers to
wear anti-vibration gloves
because the system knows that using a
high-powered drill
for repairing arc welding robots can lead
to medical conditions such as
Carpal Tunnel Syndrome
Example
Our automated planner considers
this information when generating
manufacturing plans

Coping with Complexity
Technology scales to medium-sized deployments with short execution plans
This is useful as-is for e.g. smart home environments (many services, short plans), but not for industry…
10
100
1.000
10.000
100.000
1 2 4 8 16 32 64 128 256 512 1024 2048
CalculationTime[ms]
Number of Services
Execution Plan Lenght 32 Execution Plan Length 16 Execution Plan Length 3
> 60s !
[Kovatsch, Hassan, Mayer: Practical Semantics for the Internet of Things, IoT 2015]
Execution Plan Length: 32 Execution Plan Length: 16 Execution Plan Length: 3
# Available Services
PlanningTime[ms]
log-scale
PlanningTime[ms]
log-scale

Design Space
Process-Driven Goal-Driven
Manual
Inflexible
Responsive
Automatic
Highly Flexible
Less Responsive
Mayer et al., T-ASE 13(3), 2016

Design Space
Ciortea et al., IoT 2016
Multi-Agent Systems
Mayer et al., T-ASE 13(3), 2016IFTTT, Node-RED, etc.
Intriguing Design Space!
WS-BPEL, SAP MII, etc.
OWL-S XPlan (PDDL)
Manual
Inflexible
Responsive
Automatic
Highly Flexible
Less Responsive

Design Space
Hypermedia-based Logic Planning + Multi-agent Systems
Intriguing Design Space!
[Ciortea, Mayer, Michahelles: Repurposing Manufacturing Lines on the fly with MAS for the WoT, AAMAS 2018]
Ciortea, Mayer, Michahelles, AAMAS 2018

[Ciortea, Mayer, Michahelles: Repurposing Manufacturing Lines on the fly with MAS for the WoT, AAMAS 2018]
Automatic Reasoning and Multi-agent Systems
Approach: combine agent-based strategic planning with hypermedia-based logic planning
− Build interfaces on current standardization by the WWW Consortium (W3C): Web of Things Thing Description (TD)
− Use MAS organizations to reduce the overlap of state spaces between agents
− Plans are expressed in machine- as well as human-readable form
Collaborative planning by people and software agents
− Agents can use an automated planner to derive plans
− Agents can use manually created plans
− Agents can use an automated planner to assemble plan fragments
→ Ability to select the desired level of autonomy of the system
→ Faster planning (due to constraints from/on the agent environment)
→ More flexible plans, e.g., concurrency, interweaving of plans, etc. (due to agents’ higher level of control)

Taking one step back…
Resources are integrated in the Web‘s hypermedia fabric
Logic planners reason over resource profiles
Software agents and people use logic planners to manage
hyperlinks and automatically create resource mashups
[Mayer et al.: Hypermedia to Connect them All – Autonomous Hypermedia Agents and Socio-Technical Interactions, Internet Technology Letters, 2018]
[Mayer et al.: HoloInteractions: Visualizing Interactions between Autonomous Cognitive Machines, IoT 2016]

They can also discover and integrate new resources…

…and execute mashups by traversing hyperlinks

…and execute mashups by traversing hyperlinks
And we can visualize the resulting relationships and interactions

Orchestrating Hypermedia Systems
Fundamental principle applies to any referential structure:
1. Agents suggest or create bridges between (information) silos
2. We can monitor them, and visualize these bridges
3. Implications for social studies, law, economics, and even research itself

Example: E-commerce
Autonomous agents as personal assistants for online shoppers
Logic planning to discover links between competitors‘ goods and services
Browser transparently enriches a user‘s Web experience with alternatives
„Agent-driven HATEOAS“

Example: Online Social Networks
Agents raise awareness about social media dynamics
− Confirmation bias and group polarization („echo chambers“)
− These get reflected in the hypermedia fabric (e.g., through „likes“)!
Agents supply insights to counteract misinformation
− Insert information about source/credability of content
− Insert hyperlinks to make users aware of communities with different opinions
Bridging silos: create a shared space for (political) discussion
− Example: YouTube‘s experiments
[Del Vicario et al.: Echo Chambers - Emotional Contagion and Group Polarization on Facebook. Sci. Rep. 6, 37825, 2016]
[Quattrociocchi: How does misinformation spread online? World Economic Forum Annual Meeting, 2016]

Example: Scientific Research
Agents raise awareness about isolated research communities
Agents supply insights to make researchers aware of potential bridges
Bridging silos: create a shared research space
…baby steps: analysis of how isolated researchers use social media

Conclusion
The World Wide Web is the most important distributed information system on the planet
Let’s teach agents to use and manage hypermedia systems…
− …initially, to enable flexible collaboration of machines and people in industrial environments
− …the principle applies to any domain with referential relations that are (or can be) expressed in the form of hyperlinks
− …this allows us to gain insights into the existing and modified hypermedia systems, and to visualize dependencies
Economic impact: flexible integration of siloed devices and services
Societal impact: analyze and manage interdependencies between documents, services, and people
Even more relevant – and urgent – with the progress of Ubiquitous Computing!
Let’s see if Hypermedia + MAS are up to the challenge, together!

Final note…
… we have two positions for junior researchers interested in applying MAS & SemWeb concepts in industry ;-)
… we have an open position for a senior researcher to take over a research group in this field as „area manager“
… all advertised positions at www.pro2future.at (or rather send me an email ;-))

Partners
Ecole des MINES de Saint-Étienne: Andrei Ciortea, Olivier Boissier
Inria Sophia Antipolis: Fabien Gandon
Siemens Corporate Technology: Florian Michahelles, Dominic Plangger, Mareike Kritzler, Jack Hodges
Ghent University: Ruben Verborgh
ETH Zurich Distributed Systems Group and Students: Gábor Sörös, Andreas Tschofen, Nadine Inhelder, Yassin Hassan
Karlsruhe Institute of Technology: Simon Rothfuß
Technical University of Munich: Ahmed Shafei
Massachusetts Institute of Technology: Erik Wilhelm, Josh Siegel, Sanjay Sarma
Carnegie Mellon University: Anind K. Dey
Literature
S. Mayer, D. Plangger, S. Rothfuß, F. Michahelles: UberManufacturing – A Goal-Driven Collaborative Industrial Manufacturing Marketplace. 6th International Conference on the Internet of Things, 2016
S. Mayer, R. Verborgh, M. Kovatsch, F. Mattern: Smart Configuration of Smart Environments. IEEE Transactions on Automation Science and Engineering, Vol. 13, No. 3, pp. 1247-1255, 2016
M. del Vicario et al.: Echo Chambers - Emotional Contagion and Group Polarization on Facebook. Scientific Reports 6, 37825, 2016
W. Quattrociocchi: How does misinformation spread online? World Economic Forum, 2016
S. Mayer, J. Hodges, D. Yu, M. Kritzler, F. Michahelles: An Open Semantic Framework for the Industrial Internet of Things. IEEE Intelligent Systems, Vol. 32, No. 1, pp. 96-101, 2017
A. Ciortea, S. Mayer, F. Michahelles: Repurposing Manufacturing Lines On-the-fly with MAS for the WoT. 17th International Conference on Autonomous Agents and Multiagent Systems, 2018
S. Mayer, A. Ciortea, A. Ricci, M. I. Robles, M. Kovatsch: Hypermedia to Connect them All – Autonomous Hypermedia Agents and Socio-Technical Interactions, Internet Technology Letters, 2018
Thank you for your attention!

Image Sources
http://andrewbleakley.com
http://www.webmarchand.com
http://www.patentspostgrant.com/
http://www.orkin.com
http://sweetclipart.com
http://la-matrice.org
http://learn.randolph.k12.ma.us
http://theparisreview.org
http://bestclipartblog.com
http://freepik.com
http://http://www.rugdots.com
http://www.productdose.com
http://westwoodblinds.co.uk
http://www.dolcidigio.com
http://fronius.com
http://www.ambientdevices.com
http://www.pats.ua.ac.be
http://www.vesternet.com
http://www.rovingnetworks.com
http://www.qualitymechanicalservices.com
http://chantalteakettle.org
http://blog.eeye.com
http://cartographersguild.com
http://haverford.edu
http://clker.com
http://stickers-addict.fr
http://en.wikipedia.org
http://academycomicalarts.blogspot.com
http://www.websenor.com
http://kuka.com

Autonomous Agents for Flexible Hypermedia Systems

More Related Content

What's hot (20)

Similar to Autonomous Agents for Flexible Hypermedia Systems (20)

More from Simon Mayer (9)

Recently uploaded (20)

Autonomous Agents for Flexible Hypermedia Systems