IV: Agent-Based Systems and Multi-Agent Systems Flashcards
What does Agent-Based Systems and Multi-Agent Systems mean?
All kinds of systems that include at least one agent performing tasks. The term multi-agent systems refers to systems handling several agents, often simultaneously. A key feature of agent-based systems is the concept of the environment. A single agent solves a problem by carrying out all the work by itself, whereas multi-agents solve problems by delegating tasks to the agents in the system. Agents or multi-agents can work in a centralized and decentralized environment.
What are the advantages of agent-based systems?
Handle open environment and user preferences.
Coordinate information from distributed sources
Enhance overall system performance with computational efficiency, reliability, extensibility and reuse.
Allow interconnection and interoperation of multiple existing systems. Interconnection implies that systems can communicate with each other.
What is the history behind “multi-agent systems”?
The term multi-agent system was coined during the late 1970s. The interest in these systems soared during the mid 1990s coming from an upsurge of researchers writing about the subject. Multi-agent systems are named after the many agents performing tasks. However, another interpretation is that multi-agent systems can be viewed as a loosely coupled network of software or intelligent agents that interact to carry out tasks and together solve problems that are beyond the individual capabilities of each agent.
What are the problem domains of agents?
Problem instances include the initial state and where to stop. Problem instances also include where to go when the problem is solved.
Problem space refers to the entire environment which contains a set of states and a set of operators that change the states in the problem space graph.
Problem space graph represents the problem state compared to graphs, the states are nodes, and the operators are the vertices between the nodes.
Space complexity is the maximum number of nodes in the memory.
Time complexity is the maximum number of created nodes during runtime. As these nodes are created, the computational complexity increases with the number of nodes which in turn increases the time complexity.
Branching factor denotes the possible number of actions that can be taken.
What parts are in a simple agent program?
Performance measure
Percept
Percept sequence
Behavior of agent
Actuators
What are software agents?
Software agents are entities interacting with the environment that are designed to engage in dialogues and coordinate the transfer of information, sometimes with a negotiation facility. They do not have to use intelligence which means that they perform “non-intelligent” tasks. These are used in many areas such as finding information or vaccuum cleaning.
What are intelligent agents?
Intelligent agents are agents that are capable of flexible autonomous actions in order to meet their objectives. Moreover, intelligent agents have properties, such as a learning facility, used to either father data and information or to carry out tasks. To illustrate the contents of different intelligences against, three sorts illuminate various levels of advanced agents: simple reflex agents, model-based reflex agents and goal-based agents.
What are meta-agents, negotiation agents, rational agents and mobile agents?
Meta-agents are agents built on top of software agents and intelligent agents in order to supervise these agents. They are commonly a kind of superior intelligent agent that can handle other agents and they can monitor intelligence while performing tasks and control agents’ behavior while moving between states. They can prove the fastest or best way under given circumstances.
Negotiation agents that by means of parameters can negotiate with a third party. These parameters are given by the users and are used for setting the rules of the negotiation. Applying negotiation rules to an agent can provide powerful agent systems that can negotiate automatically. The agents need a set of negotiation actions that represent the space of possible proposals
Rational agents know alternative actions and can choose the action leading to the most optimal expected outcome. Hence, the agents are reasonable, environment-sensible agents that have a good sense of judgment. They are capable of goal-directed behavior since they can independently make decisions and execute these decisions.
Mobile agents are agents with the ability to transport themselves in a physical environment or in a network, such as physical robots and Internet search-agents moving around in the environment to solve problems.
What is swarm intelligence?
Swarm intelligence is a term used for collective behavior of decentralized and self-organized systems. It constitutes a population of many simple agents, where the agents follow a couple of simple rules. There are no centralized control structures dictating how individual agents should behave and swarm intelligence systems have a certain degree of random interactions between the agents.
What are the problems of multi-agent systems?
Multi-agents systems exhibit various problems, especially in heterogeneous distributed and dynamic systems. Agents have difficulties with handling agents, also have problems with inconsistencies between different goals and have problems obtaining consistent worldview from shared information. There are certain types of characteristics such as:
Observable vs. unobservable
Collaborative vs. non-collaborative
Deterministic vs. stochastic
Discrete vs. continuous
Episodic vs. non-episodic
What are some common intelligent agents?
Utility-based agents are agents that work with happy states and unhappy states according to exactly how happy the stakes make the agents.
Learning agents initially operate in unknown environments to then become more competent than their initial knowledge alone allowed.
User agents that act on behalf of the user. ETC: web browser
Interface agents that behave as personal assistants by providing active assistance to a user.
Believable agents that can exhibit and make inferences that are appropriate to their personality via the use of an artificial character as in appropriate to their personality via the use of an artificial character.
Temporal agents that can reason about the past and the future and can make temporal commitments for the future.
What are the problems with multi-agent systems?
Agents have difficulties handling inconsistency.
Coordination problem of managing interdependencies between agents’ activities.
