9/6/2017 Outline for today’s lecture • Intelligent Agents (AIMA 2.1-2) • Task Environments Intelligent Agents & Search Problem Formulation • Formulating Search Problems AIMA, Chapters 2, 3.1-3.2 CIS 421/521 - Intro to AI - Fall 2017 2 Thinking humanly Thinking rationally Review: Acting rationally: Thinking humanly Thinking rationally Acting humanly Acting rationally Acting humanly Acting rationally Review: What is AI? rational agents Views of AI fall into four categories: • Rational behavior: doing the right thing Thinking humanly Thinking rationally • The right thing: that which is expected to maximize goal achievement, given the available Acting humanly Acting rationally information • Rational agent: An agent is an entity that perceives and acts rationally We will focus on "acting rationally“ This course is about effective programming techniques for designing rational agents CIS 521 - Intro to AI - Fall 2017 3 CIS 521 - Intro to AI - Fall 2017 4 Agents and environments Agents • An agent is anything that can be viewed as • perceiving its environment through sensors and • acting upon that environment through actuators • Human agent: • Sensors: eyes, ears, ... • Actuators: hands, legs, mouth, … •An agent is specified by an agent function f:P a that maps a sequence of percept vectors P to an • Robotic agent: action a from a set A: • Sensors: cameras and infrared range finders • Actuators: various motors P=[p0, p1, … , pt] A={a0, a1, … , ak} • Agents include humans, robots, softbots, thermostats, … CIS 421/521 - Intro to AI - Fall 2017 5 CIS 421/521 - Intro to AI - Fall 2017 6 1 9/6/2017 Agent function & program Rational agents II • The agent program runs on the physical • Rational Agent: For each possible percept architecture to produce f sequence P, a rational agent selects an action a • agent = architecture + program expected to maximize its performance measure • “Easy” solution: table that maps every possible • Performance measure: An objective criterion for sequence P to an action a success of an agent's behavior, given the • One small problem: exponential in length of P evidence provided by the percept sequence. Revised: • Rational Agent: For each possible percept sequence P, a rational agent selects an action a that maximizes the expected value of its performance measure CIS 421/521 - Intro to AI - Fall 2017 7 CIS 421/521 - Intro to AI - Fall 2017 8 Performance measure - example Rationality is not omniscience • A performance measure for a vacuum-cleaner • Ideal agent: maximizes actual performance, but agent might include e.g. some subset of: needs to be omniscient. • +1 point for each clean square in time T • Usually impossible….. • +1 point for clean square, -1 for each move — But consider tic-tac-toe agent… • -1000 for more than k dirty squares • Rationality Guaranteed Success • Caveat: computational limitations make complete rationality unachievable design best program for given machine resources • In Economics: “Bounded Rationality” “Behavioral Economics” CIS 421/521 - Intro to AI - Fall 2017 9 CIS 421/521 - Intro to AI - Fall 2017 10 Outline for today’s lecture Task environments • Intelligent Agents • To design a rational agent we need to specify a task environment • Task Environments (AIMA 2.3) • a problem specification for which the agent is a solution : to specify a task environment • Formulating Search Problems • PEAS • Performance measure • Environment • Actuators • Sensors CIS 421/521 - Intro to AI - Fall 2017 11 CIS 421/521 - Intro to AI - Fall 2017 12 2 9/6/2017 PEAS: Specifying an automated taxi driver PEAS: Specifying an automated taxi driver Performance measure: Performance measure: • ? • safe, fast, legal, comfortable, maximize profits Environment: Environment: • ? • roads, other traffic, pedestrians, customers Actuators: Actuators: • ? • steering, accelerator, brake, signal, horn Sensors: Sensors: • ? • cameras, sonar, speedometer, GPS CIS 421/521 - Intro to AI - Fall 2017 13 CIS 421/521 - Intro to AI - Fall 2017 14 PEAS: Medical diagnosis system The rational agent designer’s goal • Performance measure: Healthy patient, minimize • Goal of AI practitioner who designs rational agents: costs, lawsuits given a PEAS task environment, • Environment: Patient, hospital, staff 1. Construct agent function f that maximizes the expected value of the performance measure, • Actuators: Screen display (form including: questions, tests, diagnoses, treatments, referrals) 2. Design an agent program that implements f on a From: The New Yorker April 2017 particular architecture • Sensors: Keyboard (entry of symptoms, findings, patient's answers) CIS 421/521 - Intro to AI - Fall 2017 15 CIS 421/521 - Intro to AI - Fall 2017 16 Environment types: Definitions I Environment types: Definitions II • Fully observable (vs. partially observable): An agent's • Static (vs. dynamic): The environment is unchanged while sensors give it access to the complete state of the an agent is deliberating. environment at each point in time. • The environment is semidynamic if the environment itself does not change with the passage of time but the agent's performance • Deterministic (vs. stochastic): The next state of the score does. environment is completely determined by the current state and the action executed by the agent. • If the environment is deterministic except for the actions of other • Discrete (vs. continuous): A limited number of distinct, agents, then the environment is strategic. clearly defined percepts and actions. • Episodic (vs. sequential): The agent's experience is • Single agent (vs. multiagent): An agent operating by divided into atomic "episodes" during which the agent itself in an environment. perceives and then performs a single action, and the choice of action in each episode does not depend on any previous action. (example: classification task) (See examples in AIMA, however I don’t agree with some of the judgments) CIS 421/521 - Intro to AI - Fall 2017 17 CIS 421/521 - Intro to AI - Fall 2017 18 3 9/6/2017 Environment Restrictions for Now • We will assume environment is • Static • Fully Observable Problem Solving Agents & • Deterministic • Discrete Problem Formulation AIMA 3.1-2 CIS 421/521 - Intro to AI - Fall 2017 19 CIS 421/521 - Intro to AI - Fall 2017 20 Outline for today’s lecture Example search problem: 8-puzzle • Intelligent Agents • Formulate goal • Task Environments • Pieces to end up in order • Formulating Search Problems (AIMA, 3.1-3.2) as shown… • Formulate search problem • States: configurations of the puzzle (9! configurations) • Actions: Move one of the movable pieces (≤4 possible) • Performance measure: minimize total moves • Find solution • Sequence of pieces moved: 3,1,6,3,1,… CIS 421/521 - Intro to AI - Fall 2017 21 CIS 421/521 - Intro to AI - Fall 2017 22 Example search problem: holiday in Romania Holiday in Romania II • On holiday in Romania; currently in Arad • Flight leaves tomorrow from Bucharest • Formulate goal • Be in Bucharest • Formulate search problem You are here • States: various cities • Actions: drive between cities • Performance measure: minimize distance • Find solution • Sequence of cities; e.g. Arad, Sibiu, Fagaras, Bucharest, … You need to be here CIS 421/521 - Intro to AI - Fall 2017 23 CIS 421/521 - Intro to AI - Fall 2017 24 4 9/6/2017 More formally, a problem is defined by: Solutions & Optimal Solutions 1. States: a set S • A solution is a sequence of actions from the 2. An initial state siS initial state to a goal state. 3. Actions: a set A — s Actions(s) = the set of actions that can be executed in s, that are applicable in s. • Optimal Solution: A solution is optimal if no solution has a lower path cost. 4. Transition Model: s aActions(s) Result(s, a) sr —sr is called a successor of s —{si } Successors(si )* = state space 5. Path cost (Performance Measure): Must be additive —e.g. sum of distances, number of actions executed, … —c(x,a,y) is the step cost, assumed ≥ 0 – (where action a goes from state x to state y) 6. Goal test: Goal(s) — Can be implicit, e.g. checkmate(s) — s is a goal state if Goal(s) is true CIS 421/521 - Intro to AI - Fall 2017 25 CIS 421/521 - Intro to AI - Fall 2017 26 Art: Formulating a Search Problem Example: 8-puzzle Decide: • Which properties matter & how to represent • Initial State, Goal State, Possible Intermediate States • Which actions are possible & how to represent • States?? • Operator Set: Actions and Transition Model • Initial state?? • Which action is next • Actions?? • Path Cost Function • Transition Model?? • Goal test?? Formulation greatly affects combinatorics of search • Path cost?? space and therefore speed of search CIS 421/521 - Intro to AI - Fall 2017 27 CIS 421/521 - Intro to AI - Fall 2017 28 Example: 8-puzzle Example: 8-puzzle • States?? List of 9 locations- e.g., [7,2,4,5,-,6,8,3,1] • States?? List of 9 locations- e.g., [7,2,4,5,-,6,8,3,1] • Initial state?? [7,2,4,5,-,6,8,3,1] • Initial state?? [7,2,4,5,-,6,8,3,1] • Actions?? {Left, Right, Up, Down} • Actions?? {Left, Right, Up, Down} • Transition Model?? ... • Transition Model?? ... • Goal test?? Check if goal configuration is reached • Goal test?? Check if goal configuration is reached • Path cost?? Number of actions to reach goal • Path cost?? Number of actions to reach goal CIS 421/521 - Intro to AI - Fall 2017 29 CIS 421/521 - Intro to AI - Fall 2017 30 5 9/6/2017 Hard subtask: Selecting a state space • Real world is absurdly complex State space must be abstracted for problem solving • (abstract) State = set (equivalence class) of real world states • (abstract)