Agents in Artificial Intelligence
Artificial Intelligence study is composed of rational agents. A rational agent could be anything which make decisions, program, machine or a person. Agent carries out the actions which give the best outcome based on past and present percepts.
An AI system contains and agent and the environment on which agent perform actions. It can be many agents in the environment. An agent is anything that can be viewed as:
- Through sensors they perceives the environment
- Through actuators they acts on the environment
Examples of Agent
Software Agent
A software agent is programmed agent which has defined programs to display files on the screen, take inputs, store data.
Robotic Agent
A robotic agent is equipped with different sensors to perform in environment. Cameras are referred as sensors and the motors are referred as actuators to perform actions
Types of Agents
Based on the perceived intelligence and capability the agents are grouped into five categories:
- Simple Reflex Agents
- Model-Based Reflex Agents
- Goal-Based Agents
- Utility-Based Agents
- Learning Agents
Simple Reflex Agents
Simple Reflex Agent perceives the environment but they work only based on current perception. Perceived history is maintained by the agent but agent perform based on the condition-action rule. Condition-action rule is a rule that maps the state i.e, condition to an action. If the condition is true the action is taken else not. The agent can only work if the environment is fully observable. For simple reflex agents operating is partially observable, it is often difficult to avoid infinite loop. If can randomize his action so it is possible to avoid from infinite loops. Limitations of Simple reflex agents are:
- Confined Intelligence
- No knowledge of the unexplored parts of environment
- Rules need to be update if any change in environment occur
Model-Based Reflex Agent
Model-based reflex agent works on finding the rules who conditions meets the current situation. It can handle the partially observable environments by use of model about the world. Agent keeps track of internal state, which is adjusted by each percept and depends on the percepts history. Agent stores the current states, which describes the structure of world, to update state it requires information about:
- How world is affected by agents action
- Without agent how world evolves
Goal-Based Agents
Goal-based agents choose their decision based on how far they are from their goal. They take every action based on how to minimize the distance from goal. This allows the agent to pick the path among multiple paths which reaches a goal state. The knowledge which helps the agent to reach their goal can be easily modified, which tends these agents to be more flexible. They usually perform search and planning. Behavior of goal-based agents can be changed easily.
Utility-Based Agent
Utility-based agents used the optimal path which lead to their goal. They choose the best path which leads to their goal among multiple paths. Achieving a desired goal is not enough, to reach a goal by cheaper, safer quicker way is the optimal goal. So utility agents use all these information to select an optimal path.
Learning Agent
Learning agent in AI is the agent which has ability to learn from its past experience. Learning agent perceive the environment and keep track every information. It starts from very basic knowledge and then able to learn incrementally from their environment.
Learning agent is composed of four components:
- Learning element: Learning element is responsible for the agent to learn from environment
- Critic: Learning element gather feedback from critic which explain how far agent is performing well according to the performance
- Performance element: Responsible for selecting external action
- Problem generator: Responsible for recommending actions that will help agent to new experience.
Thanks for Reading this Article
