Artificial Intelligence
CSE 4205
Intelligent Agents
Outline
• Agents and Environments
• Rationality
• PEAS (Performance measure, Environment, Actuators, Sensors)
• Environment types
• Agent types
Intelligent Agents
Intelligent Agents
Intelligent Agents
Intelligent Agents
Four main aspects that need to be taken into consideration when
designing an intelligent agent.
• Percepts: This is the information that the agent receives
• Actions: This is what the agent needs to do or can do to achieve its
objectives.
• Goals: This is the factor that the agent is trying to achieve
• Environment: The final aspect is the environment in which the agent
will be working in. The environment in which the agent performs is
probably the most important aspect that needs to be considered as
this affects the outcome of the percepts, actions and goals.
Agents and Environment
Vacuum-Cleaner World
(Act on a single perception)
(Act on a sequence of perception)
Rational Agent
Rational Agent
Vacuum Cleaner Agent
Specifying the task environment (PEAS)
PEAS for an automated taxi driver
PEAS for a medical diagnosis system
PEAS for Interactive English tutor
PEAS for Mushroom-Picking Robot
Omniscience
• Usually when we think of omniscience, we're thinking of an attribute of
God. But in artificial intelligence, omniscience has a different definition.
• An omniscient agent knows the actual effects of its actions.
• In other words, any AI program that knows the exactly what will happen
for all its possible actions is omniscient. By this definition, a good Tic-
Tac-Toe AI is omniscient because it always knows the outcome.
• Rationality is not the same as perfection. Rationality maximizes
Expected performance , while perfection maximizes Actual
performance.
Environment types
Environment types
Fully vs Partially Observable (Cont…)
• Consider the example of Chess where each player has access to the
complete board information. Every decision is made considering the
state of the board at that time and the possible moves by the other
player. This is a ‘fully observable’ environment.
• Contrast this with Poker where players cannot anticipate the
opponent’s game as they do not have access to the opponent’s cards.
Such an environment is ‘Partially Observable’.
Environment types
Deterministic vs Stochastic
• In a Deterministic environment, the outcome is certain and can be
determined based on a specific state. However, in a Stochastic
environment, the next state cannot be predicted with certainty. Chess
can be considered as having a Deterministic environment. The
autonomous vehicle, however, has a Stochastic environment. Here,
the decision taken by the autonomous vehicle is based on the
probability of action of other vehicles.
• In a deterministic environment, any action that is taken uniquely
determines its outcome. In chess, for example, moving a pawn from
A2 to A3 will always work. There is no uncertainty.
• In a stochastic environment, there is always some level of
randomness. Any games that involve dice are good examples - you
can never be certain that a specific number will be rolled.
Environment Types
Episodic vs Sequential
• An episode = Agent’s single pair of perception & action.
• The quality of the agent’s action doesn’t depend on other episodes.
That is, every episode is independent of each other.
• Episodic environment is simpler, the agent doesn’t need to think
ahead.
• In case of sequential, current action may affect all future decisions.
- Exp: Taxi Driver and Chess.
Environment Types
Static vs Dynamic
• A dynamic environment always changes over time.
• A Dynamic environment might change while the agent is processing a
response. However, in a Static environment, the environment does
not change during the course of decision making.
- Ex: the number of people always changes in a street.
• In Chess, the environment remains unchanged (Static) during the
period a player is attempting for the next move.
Environment types
*See book Russel and Norvig (3rd Edition): figure 2.6
Agent Types
• Four basic types in order of increasing generality:
– Simple reflex agents
– Model-based reflex agents
– Goal-based agents
– Utility-based agents
Simple reflex agents
Simple reflex agents
Simple reflex agents
• Have no Memory.
• Simple-reflex agents are simple, but they turn out to be of very
limited intelligence.
• The agent will work only if the correct decision can be made on the
basis of the current percept – that is only if the environment is fully
observable.
Model-based reflex agents
Model-based reflex agents
Model-based reflex agents
Information about how the world evolves independently of the agent:
An overtaking car generally be closer than it was a moment ago.
Goal-based agents
• The agent program can combine this with information about the results of
possible actions in order to choose actions that achieve the goal
• involves consideration of the future, "what will happen if I do...?"
• Usually requires search and planning
Goal-based agents
Goal-based agents
Utility-based agents
Utility-based agents
Goal-based vs Utility-based agent
• Well Goal Based agents will use information about what they know
and their current state to see if they have accomplished what they
wanted to or not however that alone is not enough, you also need a
utility based agent to the performance of accomplishing the goal.
• For example, if you have an autonomous car that you want to take
you somewhere, once it gets to a stop and it has the ability to turn in
3 different directions, using the goal-based agent it can know which
one it should take to get to it’s destination. Using the utility-based
agent it can measure which direction might be better to get to the
destination. Maybe one is faster but dangerous and the other is
slower but safer.
