Debugging Multi-agent Systems Using Design Artifacts: - PowerPoint PPT Presentation

1 / 29

About This Presentation

Title:

Debugging Multi-agent Systems Using Design Artifacts:

Description:

Need appropriate design methodologies and tools (Prometheus) Debugging and Testing are an important part of any system development. Debugging is important. ... – PowerPoint PPT presentation

Number of Views:33

Avg rating:3.0/5.0

Slides: 30

Provided by: scie333

Category:

more less

Transcript and Presenter's Notes

Title: Debugging Multi-agent Systems Using Design Artifacts:

1
Debugging Multi-agent Systems Using Design
Artifacts

The Case of Interaction Protocols
By David Poutakidis

2
Introduction

Agents are software
Developing agents is developing software
Need appropriate design methodologies and tools
(Prometheus)
Debugging and Testing are an important part of
any system development.

3
Debugging is important.

Up to 50 of the total cost/time of development
Much of this time is spent trying to
Locate bugs
Determine the cause of a problem
Debugging problem exacerbated by the flexibility
and complexity of agent systems

4
Current Debugging Techniques

Information gathering and visualisation
Filtering applied
Color Coded messages displayed at runtime
Limitations
Too much information to digest
No process for choosing the info to display
Relies on programmer interpreting message flow
properly

5
Our Approach Make use of Design Documents

During design, documents are produced
which represent the developers description of the
system
that describe how the system should operate
that guide the programmers during coding
Design Artifacts
interaction protocols, interaction diagrams,
scenario diagrams, capability diagrams, plan
descriptors, etc

6
Central Thesis

is that the design documents and system models
developed when following an agent based software
engineering methodology can be valuable resources
during the debugging process and should
facilitate the automatic or semi-automatic
detection of errors.

7
Interaction Protocols

Capture interaction patterns between agents
Allow choices and describe all allowable
interactions
Existing protocols developed, eg FIPA (based on
AUML)

8
Use Case and Interaction Diagram
Use case 1. Customer orders book 2. Customer
specifies delivery 3. Shop ass. Obtains checks
credit details 4. Shop ass. thanks customer and
places order
9
Merchant
Bank
M1
Credit check request (CCR)
B1
Card details request (CDR)
M2
Card details (CCD)
Approve
B2
M3
Reject-1
M4
x
Fraud
M5
Reject-2
M6
10
Debugging with Interaction Protocols
Agent
Agent
Messages
Debugger
11
AUML not suitable for debugger

AUML not used internally by the debugger
AUML is not precisely defined, no formal
semantics
decouple the underlying representation to allow
for future versions of AUML or other notations
altogether.
We use equivalent Petri Nets internally
Clear formal semantics, algorithms and tools for
checking properties (liveness, deadlock)

12
Petri Nets Explained

Consists of Places (depicted as circles) and
Transitions depicted as squares.
Places and Transitions linked by arrows.
Places can contain tokens which indicate its
marking and the state the net is in.

Before Firing
After Firing
13
Converting AUML protocol to a Petri net
equivalent.
a
a
P
Q
a
a
P
P
Q
Q
14
Translating AUML Connectors
P
a
b
a
Q
P
R
b
Q
R
AUML Selection
Petri net Selection
15
CCR
M1
B1
Bank
Merchant
CDR
M2
M6
Reject-2
M1
Credit check request (CCR)
CCD
Approve
B1
B2
Card details request (CDR)
M2
Fraud
Card details (CCD)
M3
Approve
B2
Reject-1
M3
Reject-1
M4
B2
B2
x
Fraud
M5
Reject-2
M6
M5
M4
16
CCR
B1
M1
Execution of the Debugging Agent Repeat 1.
Receive message m 2. Check m against existing
protocol instances 3. Fire the Petri Net(s) 4.
Examine the Petri Net(s) for erroneous conditions.
CDR
M2
M6
Reject-2
CCD
Approve
B2
M3
Reject-1
B2
B2
M4
M5
17
CCR