AGV Simulation

Download Report

Transcript AGV Simulation

7 July 2015
AGV Simulation
Tutorial
Nelis Boucké
Alexander Helleboogh
‹#›
Overview
• Architecture
• Installation
7 July 2015
• Introduction
• Use
• Extensions
2
Simulation is crucial for
MAS applications
• AGVs (Automated Guided Vehicles)
• Advantage of using MAS
o Decentralized system behavior
7 July 2015
o Transport loads in warehouse
o Complex network of road segments and
crossroads
Robust & flexible
In dynamic & unpredictable environments
• Test before deployment!
o Decentralized systems
o No risk of damaging AGVs
• Challenge for simulation
o Test robustness & flexibility
o Dynamic test scenarios needed!
3
How do we test?
Test
Simulated environment
Deploy
Warehouse environment
7 July 2015
Agent software
4
Overview
• Architecture
• Installation
7 July 2015
• Introduction
• Use
• Extensions
5
Architecture
AGV Simulator GUI Plug-in
agentwise.
Simulation
MAP
editor
agvsimulation.ui
Runner
7 July 2015
agentwise.
agvsimulation.
AGV Agent
behaviordemo
Behavior
AGV Batch
agentwise.
Simulator
agvsimulation
XML
Eclipse Platform
Java VM
6
AGV batch simulator
Interface
Interface
Interface
7 July 2015
Deliberation
Time Model
Simulated Environment
7
Programming
AGV Behavior
• Class:
Interface
Specify your behavior here!
Package agentwise.agvsimulation.behaviordemo
7 July 2015
o AGVRandomWalkBehavior
• Classes you can use:
o AGVBehavior
The complete set of instructions to steer the AGV
robot
package agentwise.agvsimulation.agv
8
Scenario: decisions of agents
7 July 2015
sendUnicast @ T=1
driveToStation @ T=1
driveToStation @ T=2
Influences
QuickTime™ and a
TIFF (Uncompressed) decompressor
are needed to see this picture.
9
Send activity over (1,2)
7 July 2015
Scenario: decisons result in
activities
Move activity over (1,5)
Move activity over (2,6)
Activities
QuickTime™ and a
TIFF (Uncompressed) decompressor
are needed to see this picture.
10
7 July 2015
Activities interfere!
Intended outcome
Actual outcome
11
What does this imply?
• Specification of AGV programming interface
o All kinds of catastrophes can happen
Not indicated by exceptions
7 July 2015
o No post conditions in terms of what will happen
o Only specifies what an agent tries to do
• Thinking versus acting
o Agents can think and act in parallel
Agent have to wait explicitly while their AGV is driving
waitsecond()
Agents can send messages while their AGV is driving
o Agents are not notified of anything
have to rely on polling their sensors
12