Typically, agents are categorized into two categories, goal-based and model-based. Goal-based agents have knowledge about the world and perform actions derived from it. Model-based agents are similar to reflex agents, except that their knowledge is derived from percept history. Both types of agents are capable of calculating probabilities of success. They also keep track of their internal state and the state of the world.
Way to achieve a goal
The first type of agent is called the simple reflex agent. A simple reflex agent handles the problem of breaking while driving a taxi, while a utility-based agent acts based on the best way to achieve a goal. A utility-based agent is useful when the best path to accomplish a goal requires considering the best way to get there. The utility-based type can consider a variety of possible actions and decides which one is most efficient.
Finally
The goal-based agent seeks to maximize utility by choosing the best action from several possible options. A utility-based agent acts based on its utility function, and tries to maximize its output at a given state. These agents are very useful when there are multiple alternatives. This is because utility-based agents are often the best at assessing their own performance and goals. In such a case, the most cost-effective method can achieve a desired goal in the most cost-effective way.
