Download Intelligent Agents Intelligent agents Intelligent agents

Intelligent agents
■  Area on the border(s) of computer science
Intelligent Agents
(distributed systems), SE, AI, economics,
ethology, cognitive science...
■  Beware: agents do not always mean the same
across the literature!
■  Metaphoric (?!) use of cognitive notions for
describing intelligent systems
Prof. Dr. J.-J. Ch. Meyer
Utrecht University
2
1
Intelligent agents: our definition
Properties of agents
■  Software / hardware entities that display
■  Situated: acting in an environment
■  Reactive: able to react adequately to
a certain degree of autonomy / initiative,
are proactive/goal-directed
■  Mostly described in terms of having
‘mental states’ (‘strong’ notion of
agency)
■  Show informational and motivational
attitudes
(unexpected) situations arising in the
environment
■  Pro-active: setting/pursuing its own
goals
■  Social: able to co-operate with other
agents
3
4
Agents versus objects
Multi-Agent Systems (MASs)
■  objects: some autonomy (e.g. own
■  A MAS is a group of agents, together
methods)
■  agents more ‘subjects’ than ‘objects’
■  “objects do it for free, agents do it for
money (or because they want to)”
■  “agents are objects with an attitude”
with their common, shared environment
■  Primary aim: construction of flexible,
robust systems, i.e. able to cope with
unexpected events in the environment
(without crashing or displaying other
undesired behavior)
–  ‘graceful degradation’
5
6
1
(Why) MASs ?
Why MASs ? (ctd)
■  Flexibility & robustness should be
■  Besides, from an engineering
obtained via the properties of individual
agents:
perspective, the agent’s metaphor (i.e.
using agent concepts metaphorically)
helps to design and construct
complicated (distributed) systems!!
–  Reactivity
–  Pro-activeness
–  Social ability
7
8
Applications of (multi-)agent
systems
(Multi) Robot Systems
■  Autonomous Robots
■  Traffic & transport
■  Softbots (Software agents)
–  Industrial applications
–  Commercial applications
–  Medical applications
–  Entertainment
■  Space robots
■  Rescue robots
■  Robot soccer
■  Robot companions
■  …
9
Robotic System
ROBOT
Cognitive robotics
■  Cognitive Robotics (CR) concerns the
CONTROLLER
percepts
theory and implementation of robots
that reason, act and perceive in
changing, incompletely known,
unpredictable environments
commands
BODY
stimuli
10
actions
ENVIRONMENT
11
12
2
Cognitive robotics
Cognitive robotics
“Cognitive robotics aims at creating
robots - or, more generally, any kind of
largely autonomous agents - which
are endowed with high-level cognitive
functions.”
■  Cognitive robots must have higher-level
Hölldobler & Tielscher
cognitive functions that involve
reasoning about:
–  Goals, actions, time
–  When to perceive & what to look for
–  Cognitive states of other agents
–  Collaborative task execution
13
14
High-level cognitive functions
What is cognitive robotics?
■  reasoning about actions and effects
■  CR is concerned with integrating
–  Reasoning
–  Perception
–  Action
■  perceptions
■  goals and desires
■  ‘mental state’ of other agents
within a uniform theoretical and
implementation framework
Cognitive robotics : viewing robots as
intelligent agents
15
What is cognitive robotics?
16
Cognitive Robotic System
■  In other words:
ROBOT
Cognitive Robotics takes
“The robot as intelligent agent”
as a perspective:
“Robots are autonomous”
AGENT
percepts
commands
BODY
stimuli
actions
ENVIRONMENT
17
18
3
NASA explorer robots
(Multi) Robot Systems
19
(Multi) Robot Systems
20
Autonomous vehicles
Autonomous Unmanned Aerial Vehicle - Linköping
21
Linköping’s UAV project
22
Robot soccer
23
24
4
25
26
Intelligent Robot Companions
RoboCup 2014 simulation league 2D
■  Companions of human users
–  Personal assistants
RoboCup2014
•  PSA’s for ISS (NASA)
–  Ambient intelligent systems
–  e-buddies / e-partners
•  E.g. for elderly people and children
27
Applications of (M)ASystems
28
Applications of (M)AS (2)
■  Industrial Applications
–  Process control
–  Manufacturing
–  Air-traffic control
■  Commercial Applications
–  Information management
–  E-commerce
–  Business process management
■  Medical Applications
–  Patient monitoring
–  Health care
■  Entertainment
–  Games: virtual characters
–  Interactive theatre and cinema
•  Believable agents
31
32
5
Virtual characters in games
Contents lectures
(GATE subproject with TNO and UT)
■  Aim: to make more interesting games
■  Model virtual character as a ‘BDI+’ agent
■  Sub-subprojects:
–  Mental state abduction
■  Introduction
■  Philosophical foundations
■  Logical foundations of actions and
agents
•  Application of a ‘theory of mind’ for agents
•  Michal Sindlar
–  Single-Agent Logics
–  Explainable AI
■  Reasoning about action & change
•  Use BDI reasoning for letting agents in a training
environment (serious gaming) explain their actions
•  Maaike Harbers
■  Multi-Agent Logics
33
34
Literature on philosophical
foundations
Philosophical Foundations
–  P.R. Cohen & H.J. Levesque, Intention is
Choice with Commitment, Artificial
Intelligence 42, 1990, pp. 213-261 leesmap
–  A.S. Rao & M.P. Georgeff, Modelling
rational agents within a BDI-architecture,
in: Proc. KR’91, Morgan Kaufmann,
1991website
■  Dennett’s intentional stance
■  Bratman’s theory of intentions
■  Cohen & Levesque’s approach
■  Rao & Georgeff’s BDI logic
–  M.E. Bratman, Intention, Plans, and Practical
Reason, Harvard U.P., 1987*
35
36
Logical foundations of actions
and agency
Reasoning about action & change
■  Intelligent agents, agent attitudes
■  Frame problem and related problems
■  Temporal logic
■  Situation calculus, event calculus, fluent
■  Dynamic logic
calculus
■  Planning
■  Relation with non-monotonic reasoning
■  Single-Agent Logics: BDI, KARO
37
38
6