Basic Concepts

• State

describes the agent's status with respect to the environment. In the grid world example, the state corresponds to the agent's location.

• Action
• State transtion
• Policy

A policy tells the agent which actions to take at every state

• Reward

Reward is one of the most unique concepts in reinforcement learning

• Trajectories

A trajectory is a state-action-reward chain. For example, given the policy shown in Fingure 1.6(a), if the agent can move along a trajectory as follows:

$$s_1 \underset{r=0}{a_2} s_2 \underset{r=0}{a_3} s_5 \underset{r=0}{a_3} s_8 \underset{r=1}{a_2} s_9 .$$

• Returns
• episodes