Artificial Computation Intelligence

 

Components of Problem Formulation

1. Initial State - Is the starting configuration of system.
2. Actions - The permissible moves or operations that can be performed in the problem space.
3. Transition Model - Describes the result of applying specific action in a given state.
4. Goal State - Desired configuration that indicates problem has been solved.
5. Cost function - numerical cost for each action.

Task is to setup an automated traffic monitoring system that should be equipped with 10 cameras at each intersection to detect traffic violoations and detect license plates to cross verify with stolen vehicle database. What is the problem formulation for this, provide PEAS description, which agent architecture will fit this ?

PEAS Description:

Performance Measure: The system's performance is measured by its accuracy in detecting traffic violations, identifying license plates, and cross-verifying them with the stolen vehicle database. Additionally, the efficiency of the system in processing and analyzing data, as well as its reliability in real-time operation, are important performance measures.


Environment: The environment consists of intersections with traffic flow, vehicles passing through, varying lighting conditions, weather conditions, and potential occlusions of license plates due to obstructions or other vehicles. Additionally, the system interacts with databases containing information about stolen vehicles.


Actuators: The actuators of the system include the cameras installed at each intersection, which capture video footage and images of vehicles, as well as any warning signals or alerts to authorities in the case of detected violations or matches with the stolen vehicle database.


Sensors: The sensors of the system include the cameras for capturing video footage and images of vehicles, as well as additional sensors for detecting traffic flow, vehicle speed, and environmental conditions such as lighting and weather.

Types of Agent architectures

Simple Reflex Agents:

These agents act based on a direct mapping from states to actions. They do not have memory or the ability to consider future consequences. Their actions are determined solely by the current percept.

For example, in basic reactive systems like thermostat controllers where actions 

are determined by current temperature readings.

Model-Based Reflex Agents:

These agents maintain an internal model of the world based on their observations. They use this model to anticipate the effects of their actions and make decisions accordingly.

These agents are useful in environments with predictable dynamics where actions have long-term consequences. For instance, in navigation systems where anticipating traffic patterns and road conditions helps in making informed decisions.

Goal-Based Agents:

These agents have explicit goals or objectives that they strive to achieve. They consider their current state, their goals, and the available actions to select the most appropriate course of action that moves them closer to their goals.

Goal-based agents are effective in environments with clear objectives and multiple possible actions to achieve them. They are commonly used in planning and scheduling systems, robotics, and autonomous agents tasked with goal-driven tasks.

Utility-Based Agents:

These agents evaluate actions based on a utility function that quantifies how desirable or beneficial each action is in achieving their goals. They select actions that maximize their expected utility.

These agents are beneficial in environments where actions have uncertain outcomes, and decisions need to be made considering trade-offs between different objectives. They are applied in decision-making systems such as resource allocation, game playing, and economic modeling.

Learning Agents:

These agents improve their performance over time by learning from experience. They can adapt to changes in their environment and optimize their behavior 

through learning mechanisms such as reinforcement learning, supervised learning, or unsupervised learning.

Learning agents excel in environments where the relationship between actions and outcomes is not fully known or changes over time. They are suitable for tasks like autonomous driving, where the agent learns from experience to adapt to diverse driving conditions and behaviors of other vehicles.

Deliberative Agents:

These agents consider various possible actions and their consequences before making a decision. They use reasoning and planning to select the best course of action based on their goals and the current state of the environment.

Deliberative agents shine in complex environments where planning and reasoning are required to achieve goals efficiently. They are employed in systems like logistics management, where long-term planning and optimization are crucial for resource allocation and routing.

Hybrid Agents:

These agents combine elements of different architectures to leverage their respective strengths. For example, a hybrid agent may incorporate both reactive and deliberative components to balance reactive responses to immediate stimuli with more thoughtful, strategic decision-making.

Hybrid agents are valuable in environments that demand a balance between reactive responses and thoughtful decision-making. They are useful in applications like intelligent assistants, which require both quick, context-sensitive responses and strategic planning based on user preferences.

Emotion-Based Agents:

These agents incorporate emotional states into their decision-making process. Emotions can influence the agent's goals, priorities, and actions, leading to more human-like behavior in certain contexts.

Emotion-based agents find utility in environments where emotional states influence decision-making and interactions with humans. They are employed in applications like virtual assistants, therapy systems, and affective computing, where understanding and responding to human emotions enhance user experience and engagement.

Types of Environments

1. Partially Observable - Features of the environment relevant to the actions are only partially observable.
2. Fully observable 
3. Deterministic - Next state of environment is determined by current state and agent action.
4. Stochastic - Action and current state alone does not determine outcome of action but the outcomes are influenced by randomness.
5. Sequential - the current action is not just defined based on the immediate actions but also on the history of past actions. Board games is an example where past actions will be relevant.
6. If the number of states of the environment is limited it is a discrete else continuous environment.

Evaluating Heuristic design 

Heuristic design involves creating problem-solving strategies or algorithms that leverage domain-specific knowledge or rules of thumb to guide the search for solutions in complex problem spaces. Heuristics are typically used when an optimal solution is either impractical to find or computationally expensive.

Admissibility - Heuristic cost given in the problem (h) for the node is less than the actual path cost from the given node to goal node (h*).

h<=h* means admissible

Consistency Or Monotonic

Heuristic cost of reaching goal node directly from the given node (h(n))<= 

Heuristic cost of reaching the goal node from the sucessor node (h(n*))of the given node + actual cost of reaching successor node from given node (C(n,n*))

h(n)<= h(n*)+ c(n,n*) means consistent