Problem type

from Wikipedia, the free encyclopedia

In artificial intelligence there are four types of problems :

  1. The one-state problem
  2. The multi-state problem
  3. The contingency problem
  4. The exploration problem

The problem types mentioned occur mainly in connection with planning agents .

Problem types

One state problem

In the case of a one-state problem, the agent is fully aware of its own state. (The area is accessible ). The agent also knows the consequences of his actions.

Armed with this knowledge, the agent can come up with a plan that will bring him to a target state.

Multi-state problem

In the case of the multi-state problem, the agent knows the consequences of his actions. However, the agent knows not what state he is in. However, the agent is aware of which states the environment can assume.

With this knowledge, the agent can draw up a plan. In contrast to the one-state problem, the nodes of the plan do not consist of individual world states, but of all the world states that are still possible through the previous sequence of actions. Each further action reduces the set of world states in which the agent can currently be, or at least leaves them the same.

Contingency problem

The contingency problem is characterized by the uncertainty of actions.
The agent knows his current situation. He is also aware of the possible actions he can take and what they do. However, the agent cannot assume that its actions will be successful (and the agent knows that). The agent is still able to plan with uncertain knowledge .

Exploration problem

The agent does not know the current state of the world, nor what consequences the agent's actions have on his environment . Therefore the agent is forced to try in order to learn.