The Nature of Environments


The Nature of Environments

Environments are further divided into several categories. This would aid in determining the intelligence that the spy would require.

  • Full or Partial Observability? The agent's sensors do not need to pre-store any information if they have complete access. Partial access may be sensor inaccuracy or insufficient environmental data, such as limited access to hostile territory.
  • Number of Agents — A single agent environment is used for the vacuum cleaner, but for driverless taxis, each driverless cab is a different agent, resulting in a multi-agent environment.
  • Deterministic — The number of unknowns in the environment impacts the ecosystem's predictability. For example, cleaning floor space is generally predictable, and furniture stays there most of the time, while taxi driving on the road is not.
  • Deterministic — The number of unknowns in the environment impacts the ecosystem's predictability. For example, cleaning floor space is generally predictable, and furniture stays there most of the time, while taxi driving on the road is not.
  • Static — How frequently does the surrounding environment change? Is it possible for the agent to learn about the surroundings and then perform the same thing every time?
  • Episodic — A response to a precept is discrete if it is not reliant on the preceding one, i.e. it is stateless (static methods in Java). A sequential environment exists when a recent decision impacts future decisions.