CSC 411 1st Edition Lecture 4Outline of Last Lecture II. Agents and EnvironmentsIII. RationalityIV. PEASOutline of Current Lecture II. Environment Typesa. Fully vs. Partially observableb. Deterministic vs. Stochasticc. Episodic vs. Sequentiald. Static vs. Dynamice. Discrete vs. Continuousf. Single vs. MultiagentCurrent LectureFully observable vs. Partially observable- Is everything an agent requires to choose its actions available to its sensors?o If so, environment is fully accessibleo If not, parts of the environment are inaccessible Agent must make informed guesses about the worldDeterministic vs. Stochastic- Does the change in world state depend only on the current state and agent’s action?- Non-deterministic environmentso Have aspects beyond the control of the agento Utility functions have to guess at changes in the worldEpisodic vs. Sequential- Is the choice of current action dependent on previous action?o If not, then the environment is episodic (rock, paper, scissors)- In non-episodic environmentso Agent has to plan ahead: current choice will affect future actionsThese notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.Static vs. Dynamic- Static environments don’t change while the agent is deliberating over what to do- Dynamic environments do changeo So agent should/could consult the world when choosing actionso Alternately: anticipate the change during deliberation OR make decision very fast- Semidynamic: if the environment itself does not change with the passage of time but the agent’s performance score does- Exampleso Crossword: statico Poker: statico Part-picking: dynamicDiscrete vs. Continuous- A limited number of distinct, clearly defined percepts and actions (discrete) or a big range of values (continuous)Single Agent vs. Multiagent- An agent operating by itself in an environment or many agents working together- Exampleso Crossword: singleo Poker: multio Taxi driver: multio Part-picking: single- Generally, single-agent is much easier to design, but if a single-agent system cannot reach the goal, must design
View Full Document