observe environment

update internal state

evaluate possible actions

choose the best action

execute action

repeat