Intelligent Agents
E N D
Presentation Transcript
Intelligent Agents R & N Chapter 2
Rationality • “A rational agent is one that acts so as to achieve the best outcome or, when there is uncertainty, the best expected outcome.” • What does “best” mean? • What’s the outcome? • How much does it cost to get it? • How can we compute the best expected outcome? • Bounded rationality (satisficing search)
PEAS Descriptions and State Space Search • PEAS: What’s observable in the world **Given by the task • Performance Measure • Environment • Actuators • Sensors • State space search: Internal reasoning **Created by designer • States • Initial state • Goal state(s) • Successor function • Costs (of paths and/or states)
How Do They Relate? • Criminal defense lawyer Perry Mason style: • PEAS: What’s observable in the world • Performance Measure • Environment • Actuators • Sensors • State space search: Internal reasoning • States • Initial state • Goal state(s) • Successor function • Costs (of paths and/or states)
How Do They Relate? • Driving to California: • PEAS: What’s observable in the world • Performance Measure • Environment • Actuators • Sensors • State space search: Internal reasoning • States • Initial state • Goal state(s) • Successor function • Costs (of paths and/or states)
PEAS Examples Perf Envir Acts Sensors Measure Robot soccer player Internet book-shopping agent Autonomous Mars rover Mathematicians’s assistant Ask Jeeves School lunch planning
Performance Measures • Two kinds of considerations: • Where do we end up? • How much does it cost (in time, gas, or whatever) to get there? • Sometimes only one of these matters, sometimes they both do. • Examples: • Just 1 matters: • Just 2 matters: • Both 1 and 2 matter:
Is the Solution a Path or a State? Sometimes the solution is a state: But sometimes we need to output the best path that we found:
Characterizing the Environment Obs Det Epi Static Disc Agents Complexity Soccer player Defense lawyer Ask Jeeves