关键词: 强化学习, Sarsa算法, 瞬时差分算法, Q-学习算法, 函数估计

Abstract:

Reinforcement Learning does not need prior knowledge,and it gets optional policy through trial and error,its capacity of self-improving and online learning is one of the basic technologies of intelligent agent.In the paper,we firstly introduce the model and foundation of RL,then,we deeply discuss the main algorithms of RL,including Sarsa,temporal difference,Q-learning and function approximation,finally,we briefly introduce some applications of RL and some future research direction.

Key words: reinforcement learning;sarsa;temporal difference;Q-Learning;function approximation