Download Autonomous Units

Document related concepts

Knowledge representation and reasoning wikipedia , lookup

Philosophy of artificial intelligence wikipedia , lookup

Perceptual control theory wikipedia , lookup

Ecological interface design wikipedia , lookup

Machine learning wikipedia , lookup

Intelligence explosion wikipedia , lookup

Pattern recognition wikipedia , lookup

Agent (The Matrix) wikipedia , lookup

Agent-based model in biology wikipedia , lookup

Genetic algorithm wikipedia , lookup

AI winter wikipedia , lookup

Incomplete Nature wikipedia , lookup

Adaptive collaborative control wikipedia , lookup

Embodied cognitive science wikipedia , lookup

Existential risk from artificial general intelligence wikipedia , lookup

Cognitive model wikipedia , lookup

Agent-based model wikipedia , lookup

History of artificial intelligence wikipedia , lookup

Ethics of artificial intelligence wikipedia , lookup

Transcript
Artificial Life: How can it impact
engineering practices of the
future?
Cihan H. Dagli
Smart Engineering Systems Laboratory
Engineering Management Department
University of Missouri - Rolla
Rolla, MO 65409 - 0370
http://www.umr.edu/~dagli
[email protected]
ADVIS '04
1
Presentation Outline




Engineering Systems of the Future
What is Artificial Life?
Artificial Life in Engineering
Concluding Remarks
ADVIS '04
2
Recent Market Changes




Total Globalization
Increasing Production Pace
Decreasing Production Cycle Times
Migration From Mass Production to Mass
Customization
ADVIS '04
3
Engineering Systems of the
Future




Immediate Respond to Market Changes
More Sensitive to Customer Needs
Migration from Central to Distributed
Control
Autonomous and Cooperating Production
Units
ADVIS '04
4
Smart Systems

The term “smart” indicates physical
systems that can interact with their
environment and adapt to changes
through self-awareness and perceived
models of the world, based on quantitative
and qualitative information.
ADVIS '04
5
Autonomous Units
ADVIS '04
6
Autonomous Engineered
Entity
ADVIS '04
7
Autonomous Engineered
Enterprises
ADVIS '04
8
Evolutionary Color Images: Karl Sims
ADVIS '04
9
Evolutionary Color Images: Karl Sims
ADVIS '04
10
“Trajectories” of Research into
Distributed Systems
Distributed
Artificial
Intelligenc
e
Swarm
Intelligence &
Synthetic
Ecosystems
Multiagent
Systems
Population
Biology&
Ecological
Modeling
Artificial
Life
ADVIS '04
11
What is Artificial Life?

A Perspective:



It is a way of imitating Nature in order to solve
engineering problems.
It includes simulation and emulation of living
systems like plants or animals.
It tries to achieve a new understanding of
living systems, and of what is life.
http://kal-el.ugr.es/pitis.html
ADVIS '04
12
What is Artificial Life?
A Definition:
Artificial life is a field of study devoted to
understanding life by attempting to abstract
the fundamental dynamical principals
underlying biological phenomena, and
recreating these dynamics in other physical
media – such as computers – making them
accessible to new kinds of experimental
manipulation and testing.
(by Christopher G. Langton, from the preface to the Proceedings of the Workshop on Artificial Life,
February 1990, Santa Fe, New Mexico)
ADVIS '04
13
Adaptive Autonomous Agents





Agent:
A system that tries to fulfill a
set of goals in a complex, dynamic
environment.
Environment:
Adopted from http://www.rt.el.utwente.nl/agent/

It can sense the environment through

its sensors and act upon the
environment using its actuators.

Modeling Adaptive Autonomous Agents, Pattie Maes
ADVIS '04
14
Adaptive Autonomous Agents
Goal:
An agents goal can take many
different forms:
 End Goals, particular states the
agent tries to achieve
Adopted from http://www.rt.el.utwente.nl/agent/
 Selective reinforcement or reward that the
agent attempts to maximize
 Internal needs or motivations that the agent
has to keep within certain viability zones.
Modeling Adaptive Autonomous Agents, Pattie Maes
ADVIS '04
15
Agent

Autonomous


Goal-directed


Autonomous actions are directed towards the
achievement of defined tasks
Intelligent


Capable of effective independent action
Ability to learn and adapt
Cooperate

Cooperate with other agents to perform a task
ADVIS '04
16
Agent Types
Cooperate
Collaborative
Learning Agents
Learn
Smart Agents
Autonomous
Interface agents
Collaborative Agents
ADVIS '04
17
Emergent Phenomena


Emergent phenomena are those in
which even perfect knowledge and
understanding may give us no
predictive information. In them the
optimal means of prediction is
simulation. (Vince Darley, 1994)
The whole is greater than the sum of
the parts
ADVIS '04
18
Artificial Life Techniques





Agent-based modeling
Evolutionary programming
Genetic algorithms
Distributed artificial intelligence
Swarm intelligence
ADVIS '04
19
Artificial Problem Solvers:
Agent-based Modeling



Computational method where a system is
modeled as a collection of autonomous
decision-making entities that interact in
non-trivial ways.
Bottom-up modeling
Artificial social systems
ADVIS '04
20
If
<cond>
Then
<action1>
Artificial world
Else
<action2>
Inanimate agents
Observer
Animate agents
Data
Organizations of agents
ADVIS '04
Courtesy of Lars-Erik Cederman
21
Areas of Application




Flow management: evacuation, traffic,
supermarket
Markets: stock market, electronic auctions,
ISP market
Organizations: operational risk,
organizational design
Diffusion: diffusion of innovation, adoption
dynamics
ADVIS '04
22
Flow Management
Source: www.helbing.org
ADVIS '04
23
Artificial BIOWAR
detectionprivacy
Agent Location,
Demographic
& Social Network
Characteristics
Disease
Model
Geographic
Topology
Model
Environmental
Lethality
Communication
Technology
Model
Agent Model
Move
Spatially
Exposed
What If Scenario
Move
Information
Contracts Manifests
Reports
Disease Symptoms
Shared
BSS
Database
NEDSS
Compliant
Daily
Community Level
Reports
Impact Analysis
ADVIS '04
Courtesy of K. Carley, A. Yahja, B. Kaminsky
24
Artificial Problem Solvers:
Algorithms



Artificial Life tools have led to development
of many interesting algorithms that often
perform better than classical algorithms
within a shorter time.
These algorithms generally contain explicit
or implicit parallelism.
They resort to distributed agents, or to
evolutionary algorithms, or often to both.
ADVIS '04
25
Evolving Neural Networks

To develop a hybrid intelligent system –
Evolving Neural Networks (ENNs) – that
can be used in data mining, especially in
classification problems.
ADVIS '04
26
Evolving Neural Networks

Employs computational intelligence
methodologies


Neural Networks & Genetic Algorithms
Genetic algorithms have been applied to
automatic generation of neural networks




Feature selection
Adaptable topology
Customized tasks
Ensemble method
ADVIS '04
27
Optimizing a NN architecture Using GA
Genetic Algorithms
chromosomes
Evalutation of neural
network performance
fitness function:
f(x)
Translation into neural
networks
Training neural networks
ADVIS '04
28
Ensemble of ENNs
Final Decision
GA
Combining Module
GA
Evolving NN 1
Evolving NN 2
Evolving NN n
Feature Selection
Features
ADVIS '04
29
Ensemble of ENNs

ENNs meet the major requirements of a
data mining tool

Smart architecture


GA  Self-adaptable structure
Performance
Ensemble method  Accuracy
 Low complexity  Efficiency


User interaction

Objective function  Customized classification
ADVIS '04
30
Artificial Problem Solvers:
Reinforcement Learning Methods



Focus on the rational decision-making process
under uncertain environments
Agent can generate a series of actions to
influence the evolution of a stochastic dynamic
system
Underlying control problem is often modeled
as a Markov Decision Process (MDP).
ADVIS '04
31
Reinforcement Learning
Methods

What to be learned



Mapping from situations to actions
Maximizes a scalar reward or reinforcement signal
Learning


Does not need to be told which actions to take
Must discover which actions yield most reward by
trying
ADVIS '04
32
Adaptive Critic Design (ACD)


The neural control design philosophy
Algorithms are intermediate between
Direct Reinforcement and Value Function
methods, in that the “critic” learns a value
function which is then used to update the
parameters of the “actor”
ADVIS '04
33
Need for Online Hybrid Prediction
Model Derived from ACD



Fundamental drawbacks of supervised
learning-based prediction model
Uncertain volatility in real world call for
adaptive model
Reinforcement learning philosophy is
suitable tool especially when the shorttime performance of forecasting can be
obtained
ADVIS '04
34
Supervised Learning Assisted
Reinforcement Learning Prediction
Architecture for Time-Series
ADVIS '04
35
Stock Price Prediction
ADVIS '04
36
Adaptive Model Evolution
ADVIS '04
37
Artificial Problem Solvers:
Robotics

Many robotic systems are currently
being developed in the spirit of artificial
life. They are devoted to harvesting,
mining, ecological sampling etc.
ADVIS '04
38
Cooperative Behaviour & path Planning for
Autonomous Robots Using Evolutionary
Algorithm & Fuzzy Clustering
ADVIS '04
39
Alice
ADVIS '04
40
Artificial Problem Solvers:
Evolvable Systems



Different categories depending on the
complexity and purpose:
Artificial Life
Evolvable Hardware (EHW)




analog
digital (FPGAs)
Hardware design using evolution
Evolutionary Robotics
Evolving controllers for a purpose

Co-evolution of robot populations
ADVIS '04
41
Artificial Problem Solvers:
Mobile Agents
Wired network
Technical
specs
Troop
positions
Wireless
Network
Orders and
memos
George Cybenko and Bob Gray Thayer School of Engineering Dartmouth College
george.cybenko,robert.gray}@dartmouth.edu
ADVIS '04
42
Static & Mobile Agents Developed for
Small Unit Operations
Sensor Report Sent
Threat identified and Alert sent
Grapevine
Analysis agent
Sensor Field
Objectives:
• Gather information from sensor reports
• Infer additional information from object ontology
• Determine the degree of threat via fuzzy logic inference engine
• Determine recent nearby alerts using clustering
• Intelligent “push” of relevant threat data via Grapevine
Courtesy of McGrath et al
ADVIS '04
43
Artificial Problem Solvers:
Mobile Agents
Continuous data
collection
NSF 1998
KDI
Project
Operational
simulation 1
Intermittent
data
collection
Mobile agents
link weakly
coupled
distributed
components.
Data and
simulation cloud
Operational
simulation 2
Unexpected (such as
emergency relief) uses
George Cybenko and Bob Gray Thayer School of Engineering Dartmouth College
{george.cybenko,robert.gray}@dartmouth.edu
ADVIS '04
44
Multi Agent Co-operative Area
Coverage using GA




Multi Robot System
Cover Predetermined Area (Go over every
square inch)
Boundaries Marked
Minimize Time and hence Energy Efficient
ADVIS '04
45
Artificial Problem Solvers:
Swarm Intelligence

“Any attempt to design algorithms or distributed
problem-solving devices inspired by the collective
behavior of social insect colonies and other animal
societies.“
-[Bonabeau et al., 1999]-
ADVIS '04
46
Swarming Characteristics
Entities
share
common
goal
Local
Interaction
s
Autonomy
of units
Self
Organizatio
n
Simple rules
or units
Stigmergy
Distribute
d
Large
number or
efficient
size
Pulsing of
force
Flexible
and
robust
Swarming
ADVIS '04
47
Emergent- Self assembled Nest
Courtesy of Bonabeau
ADVIS '04
48
Ant Colony Optimization
1. Straight Pheromone Trail
2. Obstacle Introduced
3. Two Options are Explored
4. Shortest Path Dominates
ADVIS '04
49
Routing in Communication
Networks
ADVIS '04
50
Future Combat Systems
Courtesy of Riggs J.
ADVIS '04
51
Particle Swarm Optimization


Original intent was to simulate the
choreography of a bird flock
Best strategy to find the food is to
follow the bird which is nearest to the
food
ADVIS '04
52
PSO Initialization: Positions
and velocities
Courtesy of Maurice Clerk
ADVIS '04
53
Particle Swarm Optimization
•The best solution (fitness) particle has achieved so far
(pbest)
•The best value obtained so far by any particle in the
population (gbest)
Global
optimum
Courtesy of Maurice Clerk
ADVIS '04
54
Artificial Problem Solvers:
Synthetic Ecosystems
The synthetic ecosystems approach
applies swarm intelligence to the design of
multi-agent systems.
 The main concern of research into
synthetic ecosystems is to provide
practical engineering guidelines to design
systems of industrial strength
[Parunak, 1997] [Parunak et al., 1998]

ADVIS '04
55
Distributed Architectures for
Manufacturing

Holonic Systems



A whole individual and a part at the same time
“An autonomous and cooperative building block of a
manufacturing system for transforming, transporting,
storing and/or validating information and physical
objects”
[Christensen, 1994]
A manufacturing holon comprises a control part and
an optional physical processing part. Multiple holons
may dynamically aggregate into a single (higher-level)
holon.
ADVIS '04
56
Distributed Architectures for
Manufacturing


The application of the holonic concept to
the manufacturing domain is expected to
yield systems of autonomous, cooperating
entities that self-organize to achieve the
current production goals.
Such systems meet the requirements of
tomorrow's manufacturing control systems.
ADVIS '04
57
Concluding Remarks


Artificial Life is impacting engineering
systems through Agent-Based
architectures
Current Impact Areas:



Enterprise Integration and Supply Chain
Management
Design and Manufacturability Assessments
Enterprise Planning, Scheduling and Control
ADVIS '04
58
Concluding Remarks
Current Impact Areas:






Dynamic System Reconfiguration
Factory Control Architectures
Holonic Manufacturing Systems
Distributed Dynamic Scheduling
Commercial scheduling, routing, and force allocation
problems
Use of swarm networks to control swarm Unmanned
Aerial Vehicles (UAV), or undersea vehicles (UGV)
ADVIS '04
59