Introduction Introd ction to Intelligent S stems Systems (agents)
Outline
Agents and environments Rationality PEAS
Performance measure, Actuators, Sensors Environment,
Environment types Agent types
2
Agents agent noun: a person or thing that acts … ( (root of Latin agere, to do) g , )
3
Agents act = A t AgentFn(percept) tF ( t) sensors agent fn
actuators
4
Agents (Some definitions)
• Agent: anything that can be viewed as p g y g perceiving its g environment through sensors and acting upon through actuators • Percept: refers to the agent’s perceptual inputs at any given instant of time • Percept sequence: complete history of everything the agent has perceived th i d • Agent’s action at any instant of time depends on the percept sequence • Agent function: describes an agent’s behavior • Agent program: implementation of agent function
5
Agents and environments
The agent function maps from percept histories to actions: [f: P* A] The agent program runs on the physical architecture to produce f agent = architecture + program
Vacuum-cleaner world
Percepts: location and contents e g [A contents, e.g., [A, Dirty] Actions: Left, Right, Suck, NoOp
Rational agents
Tabulation of agent functions
Describes the agent Infinite !( for most agents) unless bound on percept sequence
Rational agent
Does the right thing Every entry in the table filled correctly
doing the i ht thing is better than doing the d i th right thi i b tt th d i th wrong thi thing what does it mean by right thing? Right action should cause the agent most successful need some ways t measure success d to
8
Tabulation o s p e age t functions abu at o of simple agent u ct o s
Percept sequence p q [A, clean] [A, dirty] [B, clean] [B, dirty] [B di t ] [A, clean], [A, clean] [A, clean][A, [A clean][A dirty] …………………………………… [A, clean], [A, clean], [A, clean] Right Action Right Suck Left Suck S k Right Suck
9
Performance measures