1.

Categories of Intelligent Agents in Artificial Intelligence

In Artificial Intelligence, an agent is defined as an entity that can perceive its environment through sensors and act upon that
environment using actuators in order to achieve specific objectives. Intelligent agents differ based on the complexity of their
decision-making processes, the type of data they use, and their ability to learn from past experiences. These agents can be
broadly categorized into five main types: Simple Reflex Agents, Model-Based Reflex Agents, Goal-Based Agents,
Utility-Based Agents, and Learning Agents. Each category has its own working principle, advantages, limitations, and
real-world applications.

Simple Reflex Agents: Simple reflex agents work on the principle of condition–action rules, meaning they respond directly
to current inputs without considering the history of past events. They are suitable for environments that are fully observable
and predictable. For example, an automatic thermostat senses the current room temperature and switches the heater ON or
OFF depending on the preset temperature. The key advantage of such agents is their fast response and straightforward
design. However, they are not suitable for complex or dynamic environments because they lack memory and cannot adapt to
changes beyond their predefined rules.

Model-Based Reflex Agents: These agents maintain an internal model of the environment, which represents aspects that are
not directly observable at the moment. The model helps them handle partially observable environments by predicting the
state of the world even with incomplete data. A good example is a self-driving car, which keeps track of other vehicles even
when they are temporarily out of sight, such as behind a large truck. The main advantage is that they can operate effectively
in more realistic situations where the environment is unpredictable. However, the performance heavily depends on the
accuracy of the model, and developing such models can be complex and computationally expensive.

Goal-Based Agents: Unlike reflex agents, goal-based agents act to achieve specific goals, which makes them more flexible
in decision-making. They evaluate various possible actions and choose the one that leads them closer to achieving their
goals. For instance, a GPS navigation system aims to guide the user to a given destination and can adapt the route if there is
traffic congestion or road closures. The main benefit is adaptability to changing conditions and objectives. However, the
effectiveness of these agents depends on having a well-defined goal, and evaluating all possible actions to find the best one
can sometimes be time-consuming.

Utility-Based Agents: Utility-based agents extend the idea of goal-based agents by introducing a utility function, which is a
numerical measure of how desirable a particular outcome is. Instead of just achieving a goal, these agents try to maximize
their “happiness” or satisfaction according to the utility function. For example, an AI-based investment system might aim to
maximize profit while also considering risk levels, thus making more balanced decisions. The advantage is the ability to
make trade-offs between competing goals and priorities. However, creating an accurate utility function that reflects all
aspects of the desired outcome can be challenging.

Learning Agents: Learning agents can improve their performance over time by learning from past experiences and adapting
to new situations. They have a learning element that allows them to modify their decision-making rules based on feedback
from their environment. For example, a spam email filter improves its accuracy as it processes more emails and learns to
identify spam patterns more effectively. The benefit is that they become smarter and more efficient with experience.
However, they may require large amounts of training data and can sometimes learn incorrect or biased patterns if the training
data is flawed.

2. Applying the PEAS Framework to a Real-World Scenario

The PEAS framework, which stands for Performance measure, Environment, Actuators, and Sensors, is a structured
approach used in Artificial Intelligence to describe and analyze the design of an intelligent agent. This model helps in clearly
defining what the agent is supposed to achieve, the conditions it operates in, and the tools it uses to perceive and act in the
environment. To understand this better, let us apply the PEAS framework to the example of an Autonomous Food Delivery
Robot.

Performance Measure:
The performance measure defines the criteria for evaluating how well the agent is performing its task. For a delivery robot,
the performance can be measured by several factors: timely delivery of food to the correct location, safe navigation without
collisions, preservation of food quality and temperature, minimal energy consumption, and overall customer satisfaction.
The better the robot performs in these aspects, the higher its success rate as an intelligent agent.

Environment:
The environment refers to everything that the agent interacts with while performing its task. In this case, the delivery robot
operates in a dynamic urban environment, which may include busy streets, pedestrian pathways, traffic signals, weather
variations like rain or heat, and unexpected obstacles such as parked cars or construction sites. The robot must be able to
adapt to these constantly changing conditions to complete its deliveries successfully.

Actuators:
Actuators are the components that enable the agent to perform actions in the real world. For a food delivery robot, this
includes wheels or motors for movement, robotic arms or trays for holding and delivering food packages, and display
screens or speakers for communicating with customers. These actuators physically carry out the decisions made by the AI
system, such as steering to avoid obstacles or stopping at the delivery point.

Sensors:
Sensors allow the agent to perceive its environment and gather the data needed for decision-making. For example, cameras
capture visual information about streets, traffic, and pedestrians; GPS modules determine the robot’s location; LIDAR
sensors detect obstacles; and temperature sensors ensure that the food remains fresh and within safe temperature limits
during delivery. These sensors provide continuous feedback to the AI system.

Integration of PEAS Components:

All four components work together to enable the robot’s operation. Sensors collect real-time environmental data, which the
AI system analyzes to decide on the best possible action based on the defined performance measures. The actuators then
execute these actions, such as adjusting speed, changing direction, or interacting with the customer. This closed-loop process
continues until the delivery is completed.

Conclusion:
By applying the PEAS framework, designers can clearly identify the goals, operating conditions, and functional
requirements of an intelligent agent. In the case of a food delivery robot, this structured approach ensures that every element
— from sensing the environment to delivering the order — is optimized for efficiency, safety, and customer satisfaction.