peas: agents and environments - nc state universitystamant/411/lectures/02-agents/peas.pdf · nc...
TRANSCRIPT
PEAS: Agents and EnvironmentsCSC 411: AI
Fall 2013
NC State University 1 / 21PEAS: Agents and Environments
What’s an agent?
Agents include humans, robots, softbots, thermostats, etc.
The agent function maps from percept histories to actions:f : P∗ → A.
The agent program runs on the physical architecture toproduce f .
NC State University 2 / 21PEAS: Agents and Environments
PEAS
We first specify the setting. Let’s design an automatedtaxi:
• Performance measure: Be safe, reach destination,maximize profits, obey laws, . . .
• Environment: Urban streets, freeways, traffic,pedestrians, weather, customers, . . .
• Actuators: steering wheel, accelerator, brake, horn. . .
• Sensors: video, accelerometers, gauges, enginesensors, keyboard, GPS, . . .
NC State University 3 / 21PEAS: Agents and Environments
PEAS
We first specify the setting. Let’s design an automatedtaxi:
• Performance measure: Be safe, reach destination,maximize profits, obey laws, . . .
• Environment: Urban streets, freeways, traffic,pedestrians, weather, customers, . . .
• Actuators: Steering wheel, accelerator, brake, horn. . .
• Sensors: Video, accelerometers, gauges, enginesensors, keyboard, GPS, . . .
NC State University 4 / 21PEAS: Agents and Environments
An aside on actuators and sensors
Some agents can modify their own actuators and sensorswith the use of tools.
• Wasps• Ravens• Dolphins• Gorillas• Human beings
NC State University 6 / 21PEAS: Agents and Environments
PEAS
A medical diagnosis system?
• Performance measure: Healthy patient, minimalcosts, no lawsuits, . . .
• Environment: Patient, hospital, pharmacy, doctors,nurses, equipment, . . .
• Actuators: Screen display (questions, tests,diagnoses, treatments, referrals) . . .
• Sensors: Keyboard (entry of symptoms, findings,patient’s answers) . . .
NC State University 7 / 21PEAS: Agents and Environments
PEAS
A medical diagnosis system?
• Performance measure: Healthy patient, minimalcosts, no lawsuits, . . .
• Environment: Patient, hospital, pharmacy, doctors,nurses, equipment, . . .
• Actuators: Screen display (questions, tests,diagnoses, treatments, referrals, . . .)
• Sensors: Keyboard (entry of symptoms, findings,patient’s answers, . . .)
NC State University 8 / 21PEAS: Agents and Environments
PEAS
How about an Internet shopping agent?
• Performance measure: Be safe, reach destination,maximize profits, obey laws, . . .
• Environment: Urban streets, freeways, traffic,pedestrians, weather, customers, . . .
• Actuators: Steering wheel, accelerator, brake, horn. . .
• Sensors: Video, accelerometers, gauges, enginesensors, keyboard, GPS, . . .
NC State University 9 / 21PEAS: Agents and Environments
PEAS
How about an Internet shopping agent?
• Performance measure: Price, quality,appropriateness, efficiency, . . .
• Environment: Current and future Web sites,vendors, shippers, . . .
• Actuators: Display to user, follow URL, fill in form. . .
• Sensors: Web pages (text, graphics, scripts. . .)
NC State University 10 / 21PEAS: Agents and Environments
Rational agents
Without loss of generality, “goals” are specifiable by aperformance measure defining a numerical value for anyenvironment history.
Rational action: whichever action maximizes the expectedvalue of the performance measure given the perceptsequence to date.
Rational 6= omniscientRational 6= clairvoyantRational 6= successful
NC State University 11 / 21PEAS: Agents and Environments
Environment properties
• Fully (vs. partially) observable: An agent’s sensorsgive it access to the complete state of theenvironment at each point in time.
• Deterministic (vs. stochastic): The next state of theenvironment is completely determined by the currentstate and the action executed by the agent.
• Episodic (vs. sequential): The agent’s experience isdivided into atomic “episodes” (in which the agentperceives then performs one action), and the choiceof action in each episode depends only on theepisode itself.
NC State University 12 / 21PEAS: Agents and Environments
Environment properties
• Static (vs. dynamic): The environment is unchangedwhile an agent is deliberating. (The environment issemidynamic if the environment itself does notchange with the passage of time but the agent’sperformance score does.)
• Discrete (vs. continuous): A limited number ofdistinct, clearly defined percepts and actions.
• Single agent (vs. multiagent): An agent operating byitself in an environment.
NC State University 13 / 21PEAS: Agents and Environments
Environment example
Crossword puzzle:
Observable:Agents:Deterministic:Episodic:Static:Discrete:
NC State University 14 / 21PEAS: Agents and Environments
Environment example
Crossword puzzle:
Observable: FullyAgents: SingleDeterministic: DeterministicEpisodic: SequentialStatic: StaticDiscrete: Discrete
NC State University 15 / 21PEAS: Agents and Environments
Environment example
Taxi driving:
Observable:Agents:Deterministic:Episodic:Static:Discrete:
NC State University 16 / 21PEAS: Agents and Environments
Environment example
Taxi driving:
Observable: PartiallyAgents: MultiDeterministic: StochasticEpisodic: SequentialStatic: DynamicDiscrete: Continuous
NC State University 17 / 21PEAS: Agents and Environments
Environment example
English tutor:
Observable:Agents:Deterministic:Episodic:Static:Discrete:
NC State University 18 / 21PEAS: Agents and Environments
Environment example
English tutor:
Observable: PartiallyAgents: Multi (why?)Deterministic: StochasticEpisodic: SequentialStatic: DynamicDiscrete: Discrete
NC State University 19 / 21PEAS: Agents and Environments
Environment example
Image analysis:
Observable:Agents:Deterministic:Episodic:Static:Discrete:
NC State University 20 / 21PEAS: Agents and Environments