1. Intelligent Agents

Agents include humans, robots, softbots, thermostats, etc.
Percept refers to the agent perceptual input at any given instant
The agent function maps from percept histories to actions:
f : P∗ → A
The agent program implements f on the physical architecture.

---

A rational agent chooses whichever action maximizes the expected value of the performance measure given the percept sequence to date（rational ̸= successful）

The environment type largely determines the agent design
The real world is (of course) partially observable, stochastic, sequential, dynamic, continuous, multi-agent

Agent types

Four basic types of agents in order of increasing generality:
– simple reflex agents
Decisions are made on the basis of the current percept only

– reflex agents with state
The internal state keeps track of relevant unobservable aspects of the en- vironment.

– goal-based agents
The goal describes desirable situations. The agent combines goal and en- vironment model to choose actions.
**Planning and search **are AI subfields devoted to building goal-based agents.

– utility-based agents
The utility function internalises the performance measure. Under uncertainty, the agent chooses actions that maximise the expected utility.

All these can be turned into learning agents

另外，Learning agents——The learning element uses feedback from the critic to modify the action selection. The problem generator suggests actions that lead to new informative experience.

Exploration vs Exploitation——A fundamental dilemma for learning agents

Exploitation: greedily uses what the agent has learnt to select the action that will, in the light of the current knowledge, have the best outcome

Exploration: Taking some other (possibly random) action to learn more, hoping to find something even better than what is currently known

Typically, a smart agent explores more in early stages than later on

Summary

1. Agents interact with environments through actuators and sensors
2. The agent function describes what the agent does in all circumstances
3. Agent programs implement agent functions
4. The performance measure evaluates the environment sequence
5. A perfectly rational agent maximizes expected performance
6. All agents can improve their performance through learning
7. PEAS descriptions define task environments(Performance measure, Environment, Actuators, Sensors)