1. Chapter 8: Multi-agents
Notes
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

2. Chapter 8: Multi-agents
Objectives
List and describe the dimensions of a multi-agent system: heterogeneity, control regime, cooperation, and goals
List and describe the axes for describing a MAS task (time, subject of action, movement, dependency)
List and describe the axes for describing a MAS collective (composition, size, communications, reconfigurability)
Compute the social entropy of a team.
Describe the use of social rules and internal motivation for emergent social behavior.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

3. The Study of Multiple Robots
Distributed
Artificial
Intelligence
Distributed
Problem
Solving
Multi-
Agent
Systems
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

4. The Study of Agency (after Stone and Veloso 2002)
Distributed
Artificial
Intelligence
How to solve problems
Or meet goals by
“divide and conquer”
Distributed
Problem
Solving
Multi-
Agent
Systems
Single computer:
How to decompose task?
How to synthesize solutions?
Divide among agents:
Who to subcontract to?
How do they cooperate?
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

5. 4 Dimensions of a Multi-agent System
Heterogeneity
Same (homogeneous) vs. different (heterogeneous)
Can be different on either software or hardware
Control Regime
Centralized vs. Distributed
Cooperation
Active (acknowledge each other) vs. Non-active (cooperation emerges, not explicit)
Communicating or non-communicating
Goals
Common goal (same, explicit) vs. Individual goal
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

6. The Ecological Niche of a Multi-Agent System
Remember….
Single Robot
Task
Environment
Agent
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

7. The Ecological Niche of a Multi-Agent System
Task
Environment
Individual Agent
Collective
emphasis
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

8. 3 Categories of Dependency
Independent
Robots don’t have to work directly or be aware of others
Dependent
Must work together for efficiency ex. Box pushing
Interdependent
Cyclic dependency ex. resupply
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

9. Chapter 8: Multi-agents
Box-Pushing
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

10. Chapter 8: Multi-agents
MAS Task Summary
Time
Fixed time task (ex. Collect as many cans in 10 minutes)
Minimum time (ex. Visit all rooms as fast as possible)
Unlimited time (ex. Patrol the building)
Synchronization required (ex. Push two buttons at same time)
Subject of Action
Object-based (e.g., robots place a single object- soccer)
Robot-based (e.g., robots place themselves- mapping)
Movement
Coverage (ex. Spread out to cover as much as possible)
Convergence (ex. Robots meet from different start positions)
Movement-to (ex. Going to a single location)
Movement-while (ex. Formation control)
Dependency
Independent (ex. Doesn’t require agents to know about others)
Dependent (ex. Task requires multiple agents)
Interdependent (ex. Agents depend on each other cyclically)
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

11. 2 Categories of Composition
Homogeneous
Heterogeneous ok
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

12. Chapter 8: Multi-agents
Case Studies
Georgia Tech 1994 AAAI Mobile Robot Competition team
Each robot hardware and software homogeneous
Reactive behaviors
Wander-for-goal
Move-to-goal
Avoid
Avoid-other-robots
Grab-trash
Drop-trash
Affordances
Orange=goal
Green=robot
Blue=trashcan ok
Dimensional score:
Homogeneous
Distributed control
Active cooperation (though minimal)
Individual goal
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

13. Example of Heterogeneous Team
USF USAR team
Robot had different hardware, software
Currently teleoperated navigation with autonomous reactive victim detection
Single goal, active cooperation
Confirm a victim with distributed sensors
Open door, “spotting” for navigation in confined spaces
Dimensional score:
Heterogeneous
Distributed control (could be central.)
Active cooperation
Single goal
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

14. Chapter 8: Multi-agents
Social Entropy
Way to measure heterogeneity of a collective
(go to board-> 4 identical, 4 marsupial)
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

15. Example of Heterogeneous Team
USC UAV/UGV team
autonomous helicopter transporting small robot
Currently teleoperated
Single goal, active cooperation
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

16. Chapter 8: Multi-agents
Control
- Centralized Control.
- Distributed Control.
centralized control
CONTROL and distributed control regimes. In centralized control, the robots communi-cate with a
central computer. The central computer distributes assignments,
goals, etc., to the remote robots. The robots are essentially semi-autonomous,
with the centralized computer playing the role of a teleoperator in a teleoperated
system.
In distributed control, each robot makes its own decisions
and acts independently.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

17. Chapter 8: Multi-agents
RoboCup
Robots must have a set of basic tactical behaviors
but may either receive strategic commands from the central computer.
Hybrid Reactive-Deliberative Paradigm,
reactive layer physically resides on the robot and the
deliberative layer resides on the central workstation.
Distributed control is more natural for soccer playing than centralized
control, because each player reacts independently to the situation. An example
mid-sized league
the robots are inherently heterogeneous.
Although they may be physically the same, each robot is programmedwith
a different role,most especially Goalie, Striker, and Defender.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

18. Chapter 8: Multi-agents
Cooperation
- Refers to how the robots interact with each other in pursuing a goal.
- Active cooperation: Acknowledging one another and working together. ‘does not necessarily mean the robots communicate
with each other.’
- Non-active cooperation: Robots individually pursue a goal without acknowledging other robots but
cooperation emerges.
cooperation - sensory capabilities of the robots.
physical cooperation, where the robots physically aid each other or interact in similar ways.
Marsupial robots are certainly a type of physical cooperation
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

19. Box Pushing: Dynamic Reconfigurability
cooperative mobility, where one robot might come over and help another
robot in trouble. Shigeo
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

20. Physically Reconfigurable Robots
Or small identical robots that hook up to form a useful robot.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

21. Chapter 8: Multi-agents
Class Exercise
Consider the case of resupply, where many multiple vehicles are in the field and a lesser number of smaller vehicles exist to carry fuel to them, return to base, and then carry more fuel out on demand. A field vehicle emits a message that it needs to be refueled. The message intensity increases inversely proportional to the amount of remaining fuel.
Describe the MAS task.
Describe the MAS collective.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

22. Chapter 8: Multi-agents
Goal
Single Goal
Individual Goals - purely reactive robots
Single goal
AAAI Compitition
SRI – Saphira
Coordinate with central workstation –
robots responsible for autonomous navigation
-----
Individual goal
Ron Arkin: group of robotic space “ants”.
set of behaviors: find-stationary-asteroid, move-to-asteroid, push-asteroidto-
home, and avoid-robots.
These behaviors give the robots individual goals, since there is
no awareness of the goals of the other team members.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

23. In the end…most popular
Homogeneous Non-communicating agents
Heterogeneous Non-communicating agents
Homogeneous communicating agents
Heterogeneous communicating agents
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

24. Chapter 8: Multi-agents
Class Exercise
Design a multi-agent team for USAR in terms of
Heterogeneity
Control
Cooperation
Goals
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

25. How to Get “Right” Emergent Behavior
Societal Rules vs. behaviors
Nerd Herd, Maja Mataric
What if homogeneous, individual goals operating in the same area?: example-- traffic and traffic jams
Motivation
ALLIANCE, Lynn Parker
What if have single goal, divided among homogeneous agents and one robot breaks?: example—cleaning up a nuclear spill
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

26. Mataric’s Nerd Herd and Social Rules
with ignorant coex-istence.
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

27. Explicit Social Rules vs. Behaviors
Societal Rules
Ignorant Coexistence
Basic reactive approach, except robots couldn’t recognize other robots
High degree of task interference
Informed Coexistence
Recognize each other PLUS simple social rule: if detect robot, stop and wait for time P; if still there, turn left then resume move to goal
Better
Intelligent Coexistence
Recognize each other PLUS behavior: repulsed by other robots concurrent with attraction to move in same direction as the majority
Best
The robots coexisted in a team, but did not have any knowledge of
each other. A robot treated another robot as an obstacle.
move-to-goal - avoid-obstacle
Slow progress – traffic jam
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

28. Chapter 8: Multi-agents
Motivation: ALLIANCE
Divide and conquer works until a robot fails then what about the failed robot’s area?
Robot A fails:
It may realize that it is not doing a good job: becomes increasingly FRUSTRATED and change behavior (give up) it is called ACQUIESCENCE
Allows other robots to help without task interference
Robot B is finished with its task
Sees that it is waiting for Robot A and becomes increasingly FRUSTRATED until it decides to help - IMPATIENCE
Goes and helps
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

29. Chapter 8: Multi-agents
Emotional Waiters
USF heterogeneous team
Robot Server, serve food, count treat removal
Robot Refiller
Problem in 1999: refiller blocked, Server stuck
Each robot has an emotional state generator
Emotions result from observed progress on task (and personality, motivation)
Waiter calls for refill, if refiller takes long time, gets impatient, begins to move towards the refiller (intercept), eventually goes to refill station
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

30. Chapter 8: Multi-agents
Emotions: Waiters
Wait
Get
Intercept Go
Home
Hurry
Refill
Refiller
Serve
Impatient
Request
Waiter
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

31. Example: Comms using ACL
Waiter
Refiller
Serve
Request
Wait
Refill
Impatient
Hurry
(achieve
:sender butler
:receiver leguin
:reply-with 1234
:language plain
:ontology waiters
:content refill )
Can use Agent Communication Languages such as KQML
Intercept
Get
Intercept
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

32. Chapter 8: Multi-agents
Cont.
Wait
Get
Intercept
Hurry
Refill
Refiller
Serve
Impatient
Request
Waiter
(achieve
:sender butler
:receiver leguin
:reply-with 1234
:language plain
:ontology waiters
:content hurry )
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

33. Chapter 8: Multi-agents
Cont.
Wait
Get
Intercept
Hurry
Refill
Refiller
Serve
Impatient
Request
Waiter
(achieve
:sender butler
:receiver leguin
:reply-with 1234
:language plain
:ontology waiters
:content meet me half way )
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

34. Chapter 8: Multi-agents
Summary
Many, cheap robots are often better than single, expensive robot
Multi-agents are generally at least reactive, sometimes hybrid deliberative/reactive
Dimensions for categorizing:
Heterogeneity, control, cooperation, and goals
(may change dynamically)
Interference is a big problem
Social rules
Emotions, Motivation
Social entropy can be used to measure heterogeneity
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents

35. Chapter 8: Multi-agents
Review Questions
What are the dimensions of a multi-agent system?
Heterogeneity, control regime, cooperation, goals
What are the four axes of a task in a collective?
time, subject of action, movement, dependency
What are the four axes of a collective?
composition, size, communications, reconfigurability
Which is more likely to fail to in the field?
a team R with 1 member of caste 1 and 5 members of caste 2
A team R with 6 members of caste 1
Introduction to AI Robotics (MIT Press), copyright Robin Murphy 2000
Chapter 8: Multi-agents