Information Technology Reference
In-Depth Information
If communicating with environmental objects, action model will adopt the proper
communication component. The control task is also part of action component.
From the discussion above, the agents can be defined as a map from
perceptive knowledge to actions. Suppose
is the perceptive set, A is the
possible actions which can be achieved by agents. The agent function
O
define the actions of all agents. The task of artificial intelligence is to design
agent programs which can realize the map from perceptive knowledge to actions.
The Skeleton of agent program is as follows:
Algorithm 14.1
Skeleton-Agent(percept) return action
Input static: memory
Method:
1.
memory
← Update-Memory(
memory
,
percept
);
2.
action
← Choose-Best-Action(
memory
);
3.
memory
← Update-Memory(
memory,action
);
4. return
action.
f
:
O
* →
A
The agents will change its memory to reflect the new percept after every
function call. An ideal rational agent hopes to achieve the best performance for
every percept.
Not every agent's action is the reflection of new conditions. Agents can also
create their new plan. In this condition, the knowledge of information provider is
useful only in specific time. This is the main difference between reactive agents
and deliberative agents.
14.3.2
Deliberative agent
Deliberative agent (or cognitive agent) is a distinct symbol model which includes
the reasoning ability about the environment and intelligent actions. It keeps the
tradition of classical artificial intelligence. It is a knowledge based system.
Environmental model is implemented in advance to form main knowledge base.
There are two problems of this architecture:
Conversion problem: how to translate the real world to correct symbol
description?
Representation/reasoning problem: how to represent the real entity and
process. How to let the agents in possession of the ability to make decision
according to reasoning the information at a limited time.
