Artificial Intelligence
Chapter 2
Intelligent Agents
1
Out Lines
1. Intelligent Agents
2.2 Agents and Environments
2.3 Acting of Intelligent Agents (Rationality)
2.4 Structure of Intelligent Agents
2.4.1 Agent Types
2.4.2 Simple reflex agent
2.4.3 Model-based reflex agent
2.4.4 Goal-based agent
2.4.5 Utility-based agent
2.5 Multi agent systems
2.6 Learning agent
2
Introduction to Intelligent Agent
An intelligent agent is a system capable of perceiving its environment, reasoning about
it, and taking actions to achieve specific goals or objectives.
In the context of artificial intelligence, intelligent agents can be either software programs,
robots, or even more abstract systems that exhibit intelligent behavior.
3
Agents and Environments
In the context of artificial intelligence and intelligent systems, "agents" and
"environments" are fundamental concepts that define the interaction between a system
and the world it operates within. Here’s a breakdown of each:
Key characteristics of agents include:
Sensors: Mechanisms to gather data about the environment. For a software agent, sensors
could be data inputs, while for a robot, they might be cameras or temperature sensors.
Actuators: Mechanisms to affect or change the environment. For a software agent, these
could be commands or notifications, while for a robot, they might be motors or servos.
4
Agents and Environments
Decision-Making: The process by which an agent decides what action to take based on
its perceptions. This can involve simple rules or complex algorithms.
Autonomy: The degree to which an agent can operate without human intervention. More
autonomous agents can make decisions and perform actions independently.
Environments
An environment is the external context or setting in which an agent operates. It includes
everything that the agent interacts with and affects the agent’s decision-making process.
Characteristics of environments include:
State: The current condition or configuration of the environment at a given time. For
example, the state of a game might include the positions of all players and objects.
5
Agents and Environments
Dynamics: How the environment changes over time, which could be due to actions taken
by the agent or external factors.
Observability: The extent to which the agent can perceive the environment.
An environment might be fully observable (where the agent has access to complete
information) or partially observable (where the agent only has limited information).
Action Space: The set of all possible actions an agent can take. The nature of the action
space can impact the agent's behavior and effectiveness.
6
Interaction Between Agents and Environments
• The interaction between agents and environments can be described in terms of:
• Perception-Action Cycle: Agents perceive the environment, process this information to
make decisions, and then take actions that affect the environment.
• This cycle continues, with the environment responding to the agent’s actions, which in
turn affects future perceptions.
• Feedback Loop: The environment provides feedback based on the actions of the agent,
which can influence future actions and decisions of the agent.
7
Interaction Between Agents and Environments
• Examples
• Autonomous Vehicles: The vehicle (agent) uses sensors (cameras, lidar) to perceive the
environment (road conditions, other vehicles) and takes actions (steering, braking) to
navigate safely.
• Chatbots: The chatbots (agent) perceives user inputs (text or voice) and responds based
on its programmed algorithms and context (environment) of the conversation.
8
Acting of Intelligent Agents (Rationality)
In the context of intelligent agents,
"rationality" refers to the concept of making decisions that maximize an agent’s
performance or achieve its goals based on the information available.
A rational agent acts to optimize its outcomes, given its knowledge and resources. Here’s
a closer look at rationality and how it applies to intelligent agents:
9
Acting of Intelligent Agents (Rationality)
Rationality in Intelligent Agents
1. Rational Behavior:
• Goal-Oriented: A rational agent acts in a way that it believes will best achieve its
goals or objectives.
• This means the agent’s actions are guided by a predefined set of goals and
preferences.
• Optimal Decision-Making: Rational behavior involves choosing actions that are
expected to lead to the best possible outcome.
• This involves evaluating potential actions and their consequences and selecting the
10
one that maximizes expected utility or value.
Acting of Intelligent Agents (Rationality)
Rationality in Intelligent Agents
2. Performance Measures:
• Performance Metrics: The effectiveness of an agent is often measured using
performance metrics specific to its task.
• For instance, the performance of a chess-playing agent might be measured by the number
of games won.
11
Acting of Intelligent Agents (Rationality)
Rationality in Intelligent Agents
3. Knowledge and Information:
• Perception: Rational agents must accurately perceive their environment to make
informed decisions.
• This involves processing sensory input and interpreting it correctly.
12
Acting of Intelligent Agents (Rationality)
Rationality in Intelligent Agents
4. Decision-Making Process:
• Planning: Some agents use planning algorithms to determine a sequence of actions that
will lead to the desired goal.
• Planning involves evaluating the possible future states and choosing a path that
maximizes the expected benefit.
13
Acting of Intelligent Agents (Rationality)
Examples of Rational Agents:
Self-Driving Cars: These cars are designed to maximize safety and efficiency.
They make decisions based on a combination of sensory data, maps, and pre-defined
rules of the road to navigate traffic and avoid obstacles.
Game AI: In strategy games, AI agents are designed to make moves that maximize their
chances of winning.
They evaluate possible moves and their outcomes to choose the most advantageous
actions.
Financial Trading Algorithms: These agents analyze market data to make trading
decisions that maximize profit.
They use algorithms to process large volumes of data and predict market trends.
14
Acting of Intelligent Agents (Rationality) Summary
Rationality in intelligent agents is about making decisions that are expected to best
achieve the agent’s goals given its knowledge and constraints.
It involves optimizing actions, utilizing performance measures, adapting to new
information, and sometimes using learning and heuristics to improve over time.
Rational agents strive to act in a way that maximizes their expected utility or
performance, taking into account the complexities and limitations of their environment.
15
Structure of Intelligent Agents
The structure of intelligent agents can be categorized into various types based on their
complexity and decision-making capabilities.
Agent Types:
Simple Reflex Agents
Model-Based Reflex Agents:
Goal-Based Agents:
Utility-Based Agents
16
Structure of Intelligent Agents
• Intelligent agents can be classified based on their decision-making processes and internal
structures. The primary categories are illustrated in detail below:
• Simple Reflex Agents: Operate using a set of condition-action rules.
• They react to specific stimuli in a straightforward manner.
• Model-Based Reflex Agents: Extend simple reflex agents by maintaining an internal
model of the world to handle partial Observability and to account for changes.
• Goal-Based Agents: Go beyond simple responses and internal models by considering
goals or objectives. They plan actions to achieve these goals.
• Utility-Based Agents: Make decisions based on a utility function that quantifies the
desirability of different states. They aim to maximize their overall satisfaction or utility.
17
Structure of Intelligent Agents
Simple Reflex Agent Characteristics:
Condition-Action Rules: Operates based on a set of predefined rules that map specific
conditions to actions. These rules are typically of the form: "If condition X is true, then
perform action Y."
Immediate Response: Reacts immediately to environmental stimuli without considering
the history of past interactions or the state of the world beyond the current conditions.
Example:
Thermostat: A thermostat that turns the heating system on when the temperature falls
below a certain threshold and off when it exceeds another threshold. 18
Structure of Intelligent Agents
Advantages:
• Simplicity: Easy to design and implement.
• Efficiency: Fast response due to straightforward decision rules.
Limitations:
• Limited Flexibility: Cannot handle situations not covered by the predefined rules.
• No Internal State: Lacks the ability to manage situations that require memory or
understanding of past actions.
19
Structure of Intelligent Agents
Advantages:
• Simplicity: Easy to design and implement.
• Efficiency: Fast response due to straightforward decision rules.
Limitations:
• Limited Flexibility: Cannot handle situations not covered by the predefined rules.
• No Internal State: Lacks the ability to manage situations that require memory or
understanding of past actions.
20
Structure of Intelligent Agents
Model-Based Reflex Agent Characteristics:
Internal Model: Maintains an internal representation of the world’s state. This model
helps the agent keep track of important information and context beyond immediate
perceptions.
Partial Observability: Can handle environments where the agent doesn’t have full
visibility. The internal model helps in inferring the unseen parts of the environment.
Example:
Robot Vacuum Cleaner: A vacuum cleaner that keeps track of the areas it has already
cleaned and avoids those areas. It also uses this internal model to navigate around
obstacles and plan its cleaning path.
21
Structure of Intelligent Agents
Advantages:
• Improved Flexibility: Can handle a wider range of situations than simple reflex agents.
• Contextual Awareness: Better at dealing with environments where actions have longer-
term effects or require a history of past interactions.
Limitations:
• Complexity: Requires the maintenance and updating of the internal model, which can
add to computational complexity.
• Resource Usage: More resource-intensive compared to simple reflex agents.
22
Structure of Intelligent Agents
Goal-Based Agent Characteristics:
• Goal-Oriented: Operates based on achieving specific goals or objectives. The agent
decides on actions by considering how they will contribute to reaching the goal.
• Planning: Uses planning mechanisms to determine a sequence of actions that will lead to
goal achievement.
• Example:
• Navigation Systems: GPS systems that provide directions based on the goal (destination)
and current location, considering the best route to reach the destination.
23
Structure of Intelligent Agents
Advantages:
• Versatility: Can handle a wide range of tasks and scenarios by adjusting goals and plans.
• Future-Oriented: Considers the long-term effects of actions in the pursuit of goals.
Limitations:
• Complexity of Planning: Requires algorithms to generate and evaluate plans, which can
be computationally expensive.
24
Structure of Intelligent Agents
Utility-Based Agent Characteristics:
• Utility Function: Uses a utility function to measure the desirability of different states.
Actions are chosen to maximize the agent’s overall satisfaction or utility.
• Preference Handling: Can handle multiple conflicting goals by assigning different
utilities to different states and balancing them.
• Example:
• Financial Trading Algorithms: Algorithms that maximize profit by evaluating the utility
of different trading strategies and making decisions that are expected to yield the highest
return. 25
Structure of Intelligent Agents
Advantages:
• Optimal Decision-Making: More sophisticated in handling trade-offs and achieving the
best overall outcomes based on the utility function.
• Flexible: Can adapt to various criteria and preferences, making it suitable for complex
decision-making environments.
Limitations:
• Computational Complexity: Determining and maximizing utility can involve complex
calculations and optimization.
26
Structure of Intelligent Agents Summary
• Simple Reflex Agents: React to specific conditions with predefined actions.
• Model-Based Reflex Agents: Use an internal model to handle partial Observability and
maintain context.
• Goal-Based Agents: Plan actions to achieve specific goals, considering the impact of
actions on future states.
• Utility-Based Agents: Make decisions based on a utility function to maximize overall
satisfaction or benefit.
27
Multi-agent systems (MAS)
Multi-agent systems (MAS) consist of multiple interacting intelligent agents, which
work together or compete within a shared environment to achieve their individual or
collective goals.
Types of Multi-Agent Systems:
Cooperative Multi-Agent Systems
Competitive Multi-Agent Systems
Mixed-Motive Multi-Agent Systems
28
Multi-agent systems (MAS)
Cooperative Multi-Agent Systems:
Goal: Agents work together to achieve a common objective or shared goals.
Coordination: Requires mechanisms to coordinate actions and share information.
Example: Robotic teams performing search and rescue operations where each robot has a
role and they must work together to find and rescue individuals.
29
Multi-agent systems (MAS)
Competitive Multi-Agent Systems:
• Goal: Agents compete against each other to achieve individual goals, often in a shared
environment with limited resources.
• Interaction: Strategies involve competition and possibly negotiation or conflict
resolution.
• Example: Trading algorithms in financial markets where each algorithm seeks to
maximize its own profit at the expense of others.
30
Multi-agent systems (MAS)
Mixed-Motive Multi-Agent Systems:
Goal: Agents have both cooperative and competitive elements. They may need to work
together in some aspects while competing in others.
Interaction: Requires balancing cooperation and competition.
Example: Multiplayer online games where players may form temporary alliances but
ultimately compete for individual victory.
31
Learning Agent
A learning agent is an intelligent agent that improves its performance over time through
experience.
Learning agents adapt their behavior based on past interactions with their environment to
better achieve their goals.
The concept of learning is central to making agents more effective and flexible in
dynamic or uncertain environments. Here’s a detailed overview of learning agents:
32
Types of Learning Agents
Supervised Learning Agents:
• Learning Process: Learn from examples or labeled data provided by a teacher or
external source.
• The agent is trained with input-output pairs to generalize and make predictions or
decisions on new, unseen data.
• Example: A spam filter that learns to classify emails as spam or not spam based on a set
of labeled examples.
33
Types of Learning Agents
Unsupervised Learning Agents
• Learning Process: Learn patterns and structures from unlabeled data without explicit
feedback.
• The agent tries to find hidden patterns or groupings in the data.
• Example: A clustering algorithm that groups similar customer data into segments for
targeted marketing.
34
Types of Learning Agents
Reinforcement Learning Agents
• Learning Process: Learn through trial and error by interacting with the environment.
• The agent receives rewards or penalties based on its actions and adjusts its behavior to
maximize cumulative rewards.
• Example: A game-playing agent that learns strategies by playing the game multiple
times, improving its performance based on the outcomes of previous games.
35
Types of Learning Agents
Semi-Supervised Learning Agents
Learning Process: Combine a small amount of labeled data with a large amount of
unlabeled data to improve learning accuracy. This approach is useful when labeled data is
scarce but unlabeled data is abundant.
Example: A system that improves object recognition in images by leveraging a few
labeled examples and many unlabeled images.
36
Types of Learning Agents
Self-Supervised Learning Agents
• Learning Process: Learn by generating pseudo-labels or intrinsic rewards from the data
itself. The agent uses part of the data to predict other parts, creating its own training
signals.
• Example: A language model that predicts missing words in a sentence, using the context
provided by other words in the sentence as a learning signal.
37
Types of Learning Agents Summary
Learning agents are intelligent systems that enhance their performance by learning from
experience.
They can use various learning paradigms, such as supervised, unsupervised,
reinforcement, semi-supervised, and self-supervised learning, to adapt and improve over
time.
The development and application of learning agents involve addressing challenges related
to data, exploration, generalization, and computational resources, making them versatile
tools in a wide range of applications.
38
@Z End
? Welcome
39