Artificial Intelligence
CMP 346
Unit:2 Intelligent Agents
Er. Nirmal Thapa
Email: nirmalthapa@[Link]
Lumbini Engineering College
Pokhara University
Intelligent Agents
§ An agent is anything that can be viewed as perceiving its environment through sensors
and acting upon that environment through actuators.
§ An autonomous entity which acts upon an env using sensors and actuators for achieving
goals
§ Agents interact with environment through sensors and actuators.
Ref. Book: AI by Russel and Norvig
Intelligent Agents
§ What do you mean, sensors/percepts and effectors/actions?
For Humans
§ Sensors: Eyes (vision), ears (hearing), skin (touch), tongue (gestation), nose
§ Percepts:
§ At the lowest level – electrical signals from these sensors
§ After pre-processing – objects in the visual field (location, textures, colors, …), auditory
streams (pitch, loudness, direction), …
§ Effectors: limbs, digits, eyes, tongue, …..
§ Actions: lift a finger, turn left, walk, run, carry an object, …
Intelligent Agents
§ What do you mean, sensors/percepts and effectors/actions?
A more specific example: Automated taxi driving system
§ Percepts: Video, sonar, speedometer, odometer, engine sensors, keyboard input,
microphone, GPS, …
§ Actions: Steer, accelerate, brake, horn, speak/display, …
§ Goals: Maintain safety, reach destination, maximize profits (fuel, tire wear), obey laws,
provide passenger comfort, …
§ Environment: Urban streets, freeways, traffic, pedestrians, weather, customers, …
Intelligent Agents
The vacuum-cleaner world: Example of Agent
§ Environment: square A and B
§ Percepts: [location and content] E.g. [A, Dirty]
§ Actions: left, right, suck, and no-op
Intelligent Agents: Performance measure
The concept of rationality
§ A rational agent is one that does the right thing.
§ Every entry in the table is filled out correctly.
What is the right thing?
§ Right action is the one that will cause the agent to be most successful.
§ Therefore we need some way to measure success of an agent.
§ Performance measures are the criterion for success of an agent behavior.
§ The performance measure evaluates the behaviour of the agent in an environment.
§ It defines the agent’s degree of success.
§ E.g., performance measure of a vacuum-cleaner agent could be amount of dirt cleaned up,
amount of time taken, amount of electricity consumed, amount of noise generated, etc.
§ It is better to design Performance measure according to what is wanted in the environment
instead of how the agents should behave.
Intelligent Agents: Performance measure
§ It is not easy task to choose the performance measure of an agent.
§ For example if the performance measure for automated vacuum cleaner is “The amount of
dirt cleaned within a certain time” Then a rational agent can maximize this performance by
cleaning up the dirt , then dumping it all on the floor, then cleaning it up again , and so on.
§ Therefore “How clean the floor is” is better choice for performance measure of vacuum
cleaner.
What is rational at a given time depends on four things:
• Performance measure,
• Prior environment knowledge,
• Actions,
• Percept sequence to date (sensors).
Definition: A rational agent chooses whichever action maximizes the expected value of the
performance measure given the percept sequence to date and prior environment knowledge.
Environment
§ To design a rational agent we must specify its task environment.
§ Task environment means: PEAS
§ description of the environment:
§ Performance
§ Environment
§ Actuators
§ Sensors
Example: Fully automated taxi:
§ PEAS description of the environment:
§ Performance: Safety, destination, profits, legality, comfort
§ Environment: Streets/freeways, other traffic, pedestrians, weather,, …
§ Actuators: Steering, accelerating, brake, horn, speaker/display,…
§ Sensors: Video, sonar, speedometer, engine sensors, keyboard, GPS, …
Types of Agent
1. Simple reflex agent:
§ The simplest agents
§ Take decisions on the basis of the current percepts and ignore the rest of the percept history
§ These agents only succeed in the fully observable environment
§ Works on condition-action rule, which means it maps the current state to action
§ It acts according to a rule whose condition matches the current state, as defined by the
percept.
Types of Agent
2. Model based reflex agent:
§ Can work in a partially observable environment, and track the situation.
§ Has two important factors:
§ Model: It is knowledge about “how things happen in the world”, so it is called model based
§ Internal state: it is a representation of the current state based on percept history
§ Have the model, “which is knowledge of the world” and based on the model they perform
actions.
§ Keeps track of the current state of the world using an internal model and then choose an
action
Types of Agent
3. Goal based agent:
§ The knowledge of the current state env is not always sufficient to decide for an agent to what to do
§ Agent need to know its goal which describes the situations. They choose an action, so that they can
achieve the goal
§ They may have to consider a long sequence of possible actions before deciding whether the goal is
achieved or not.
Ref. Book: AI by Russel and Norvig
§ Keeps track of the world state as well as a set of goals it is trying to achieve and choose an action
that will lead to the achievement of its goals.
Types of Agent
4. Utility based agent:
§ The best way to achieve the goal
§ Useful when there are multiple possible alternatives, and an agent has to choose in order to
perform the best action
§ The utility function maps each state to a real number to check how efficiently each action
achieves the goals
Characteristics of Intelligent Agent (IA)
§ Must learn and improve through interaction with the environment
§ Must adapt online and in the real time situation
§ Must learn quickly from large amounts of data
§ Must accommodate new problem solving rules inclemently
§ Must have memory which must exhibit storage and retrieval capacities
§ Should be able to analyze self in terms of behaviour, error and success
Learning Agents
Agent Programs Origins:
• Early idea by Turing (1950): program intelligent machines manually.
• Turing concluded that building learning machines and teaching them is more efficient.
• Modern AI often prefers learning-based systems for advanced performance.
Advantages of Learning Agents:
• Operate effectively in unknown environments.
• Enhance competence beyond initial knowledge.
• Suitable for all agent types (model-based, goal-based, utility-based, etc.).
Key Components of Learning Agents
1. Performance Element:
• Takes percepts and decides actions.
• Represents the agent's operational part.
2. Learning Element:
• Improves performance using feedback.
• Depends on feedback from the critic.
• Adjusts the performance element for better future actions.
3 .Critic:
• Evaluates performance based on a fixed performance standard.
• Provides feedback for learning since percepts alone lack evaluation context.
4. Problem Generator:
• Suggests exploratory actions for new experiences.
• Encourages long-term improvement by exploring beyond optimal immediate actions.
Learning Agents
• Learning agents refine their models to better align with reality.
• Example: Automated taxi learns braking effectiveness under varied conditions.
• Simple, slightly inaccurate models may sometimes be computationally preferable.
• External standards provide information for reflexes or utility adjustments.
• Example: A taxi driver learns to avoid violent maneuvers after losing tips due to
passenger discomfort.
• Observing human reactions helps update the agent's utility function.
• Example: A taxi learns not to honk continuously based on negative human responses.
• Learning is about modifying agent components to align with feedback.
• Improves the agent's overall performance through systematic updates.
Key Components of Learning Agents
Key Components of Learning Agents
1. Performance Element:
• Takes percepts and decides actions.
• Represents the agent's operational part.
2. Learning Element:
• Improves performance using feedback.
• Depends on feedback from the critic.
• Adjusts the performance element for better future actions.
3 .Critic:
• Evaluates performance based on a fixed performance standard.
• Provides feedback for learning since percepts alone lack evaluation context.
4. Problem Generator:
• Suggests exploratory actions for new experiences.
• Encourages long-term improvement by exploring beyond optimal immediate actions.
Questions ?
Thank you !
