An agent is best described as:
- A pro-active, autonomous entity.
- An intelligent program.
- An episodic, real-time entity.
- A reactive, dynamic entity.
- An object.
The study of Multiagent Systems is generally divided into two separate
areas of interest:
- Software and Robotic agents.
- Intelligent and Reactive agents.
- Software and e-commerce agents.
- Knowledge representation and behavioral studies.
- Object-oriented and functional agents.
An inaccessible environment is:
- One where an agent cannot obtain a complete view of the environment.
- One where only some agents can move around.
- One that does not repeat itself (i.e., its random).
- One where the variables have continuous values.
- One that cannot be found.
The difference between an agent and an object is
- An agent can refuse to provide a service.
- Agents have more member functions.
- Agents are not implemented using objects.
- There is no difference.
- Agents are not implemented using an object-oriented language.
We can formally describe an agent as a mapping between states S, and
actions A. An agent with state is formally described as (let I
represent the set of states and P the set of perceptions).
- A mapping from S to P, a mapping from I to A, and a mapping from IxP to I.
- A mapping from S to P, a mapping from P to I, and a mapping from I to A.
- A mapping from S to I, and a mapping from I to A.
- A mapping from S to A, and a mapping from I to P.
- A mapping from S to I, a mapping from I to P, and a mapping from P to A.
The four most common types of agent architectures are
- Logic-based, reactive, BDI, layered.
- Logic-based, subsumption, reactive, BDI.
- Logic-based, BDI, goal-driven, subsumption.
- Subsumption, reactive, state-based, goal-driven.
- State-based, stateless, episodic, non-episodic.
The subsumption architecture incorporates a set of reactive behaviors and
- A binary inhibits relation.
- A subsuming tree organization.
- A set of beliefs and desires.
- A behavior to state mapping.
- A behavior-choosing function.
In a BDI architecture, the agent's goals are given by
- The desires.
- The beliefs.
- The intentions.
- The current activated intentions.
- A mapping from the current intentions to the desires.
Which one of the following is not a reason for using a multiagent approach.
- Multiagent systems are faster than centralized methods.
- Information is geographically distributed.
- The system has many components.
- The system has do deal with huge amounts of content.
- The system must encompass a broad scope.
Coordination can be separated into
- Cooperation and competition.
- Distributed and centralized.
- Negotiation and competition.
- Electronic and instantaneous.
- Negotiated and pre-established.
FIPA is concerned with agency as it relates to
- Autonomy and communal integration.
- Autonomy and intelligence.
- Linguistic ability.
- Agent architecture design.
- Communication protocols.
FIPA's main product is
- Published specifications.
- Software agents.
- Agent architectures.
- Multiagent systems.
- A website.
According to FIPA's message transport specifications, the main parts of a message are
- An envelope plus a body.
- A performative and content.
- The first one and the last one.
- The sender and the receiver.
- The ontology and the language.
FIPA's message transport specifications say that messages can be transported by
- IIOP, HHTP, WAP, among others.
- HTTP only.
- The FIPA ACL.
- A transport agent.
- The Agent Management System.
FIPA's Agent Management System provides a "yellow pages" service which is implemented by the:
- Directory facilitator
- Yellow Pages agent.
- Interaction Protocol.
- Transport Service.
Which one of the following is not an element in the FIPA-ACL message structure.
If an agent receives a FIPA-ACL message that it does not understand, it must reply with the performative:
If Dexter wanted to tell Dee-Dee about his new Laboratory using the
Binary language and obeying the FIPA-ACL standard. He would have to
say something like:
- (tell :sender Dexter :receiver Dee-Dee :language Binary :content (10101000100101011))
- (tell(Dexter, Dee-Dee), about(10101010100010101), language(Binary))
- <tell>Dexter Dee-Dee 1000101010110</tell>
- (tell :sent-by Dexter :received-by Dee-Dee :using Binary :content (1010101010110010010011))
The performatives that can be used in FIPA-ACL:
- Are all defined in a FIPA specification.
- Are open to interpretation by the users.
- Are meant to be added to as people start using the system.
- Are less than 10.
- Do not have precise semantic meaning.
If you wanted to create a FIPA-compliant agent that could join an
existing multiagent system, your agent would be forced to:
- Implement any interaction protocol that it claims to understand.
- Make it smart enough so it understands all the messages the agents send it.
- Implement all the interaction protocols that the other agents use.
- Implement your own interaction agent.
- Ask permission from all the existing agents in order to join.
In AUML, a set of messages only one of which should be sent, is represented using
- A diamond with an X in it.
- A heavy horizontal bar.
- An unfilled diamond.
- A series of arrows.
- A diamond with a heavy bar across it.
A FIPA Interaction Protocol can be defined as happening between
- Agents and roles.
- Agents and humans.
- The AMS and the DF.
- Concrete and Abstract agents.
In AUML, a parameterized protocol is represented using
- A dashed rectangle.
- A black diamond.
- An unfilled diamond.
- A diamond with an X in it.
- A thick black line.
In a successful application of FIPA's Contract Net protocol, the set
of performatives sent by both agents are (in order):
- cfp, propose, accept-proposal, inform-done
- announce, propose, accept, inform-done
- cfp, reply, accept-proposal, inform-done
- cfp, bid, inform-done
- accept-proposal, inform-done
Why does the auctioneer send a reject-proposal to the bidder in the
FIPA English Auction IP?
- Because otherwise the bidder would not know if his bid is currently the highest.
- He does not. In an English auction only the final highest bid wins, so everyone else knows they lost.
- Because its part of the Contract-net protocol.
- Because it is the signal that the bidder should leave the auction.
- Because the agents' bids are sent in asynchronously.
The Semantic Web vision will be accomplished by first
- Adding semantic tags to Web data.
- Developing new agent communication languages.
- Creating a new language to replace XML.
- Implementing intelligent software agents.
- Creating a master ontology.
An XML element is
- <name element="value">
- XML does not have elements.
An XML DTD can not be used to specify
- The number of times an element can appear in a document.
- Which tags can be used.
- Which attributes can be used.
- Which elements have which attributes.
- The possible ways to nest elements.
Some of the basic modeling primitives of RDF Schema include
- class and subclass-of
- name and date
- elements and attributes
- start and end
- cfp and propose
DAML is a language for
- Describing ontologies.
- Agent communication.
- Describing agent services.
- Replacing RDF.
- Extending FIPA-ACL.
When using DAML+OIL, we usually also use tags from
- RDF Schema
- XML Semantic Tags
- XML Class Extensions
- RDF Ontology Declaration
- RDF Member Triples
Which one of the following statements is the DAML representation for
the idea that if x's parent is y, then y is x's child.
The relationships between XML, RDF, and DAML+OIL are:
- XML provides the basic syntax, RDF adds some tags to XML, DAML+OIL add some tags to RDF.
- RDF extends XML by adding triples, DAML changes the structure of RDF to support ontologies.
- Each one changes the syntax of the previous language in order to add more functionality.
- XML is the syntax, RDF the semantics, and DAML+OIL provides the inferencing abilities.
- XML and RDF are more in the same Semantic Web family, while DAML+OIL comes from the KQML family.
The difference between a reactive architecture and a subsumption architecture is that
- The subsumption architecture is a type of reactive architecture.
- The subsumption architecture contains an internal state.
- The subsumption architecture is based one goals and beliefs.
- Reactive architectures contain an internal state.
- A subsumption architecture is much faster.
In the BDI architecture, intentions do not
- Determine the most likely goals.
- Drive means-ends reasoning.
- Constrain future deliberation.
- Persist over time.
- Influence beliefs upon which future practical reasoning is based.
A Blackboard system is based on the metaphor of
- Many experts writing on a common blackboard.
- A professor teaching a class using a blackboard.
- Information that can be quickly broadcast and erased.
- A large common knowledge base.
- A common communications language.
One problem with the contract-net protocol is that
- The agents that are best equipped to handle a
particular task might be busy at the time it is
- It is very time consuming to send a cfp to all the agents.
- It is often not clear which agent won the cfp.
- It is too complicated to implement for smaller agents.
- It is not well-defined.
A reactive agent architecture is implemented in an object-oriented
language by using (HINT: remember the UML diagram):
- A Behavior class with
switch statement where each case checks on particular aspects of the world state.
- A Goal, Behavior, and Plan classes, along with a
subsumes binary relation.
- Any number of classes, where the classes "above" inhibit the ones "below".
- A bucket-brigade algorithm along with the Singleton design pattern.
In multiagent systems, coherence is
- How well a system behaves as a unit.
- How well the agents communicate with each other.
- The degree to which all agents are implemented using the same design principles.
- The lack of extraneous agents.
- The ability to document an agent's behavior.
Weiss, Section 2.2.1.
Speech act theory views human natural language as
- Insufficient for agent communications
Weiss, Section 2.2.5