cumulative reward in RL AGENT block.

I understand cumulative reward means total reward of one episode.

I set my reward like below,

reward : psi>40 ------- -1

psi<10 ------- +10

blus line is psi, yellow line is total reward,

but,

In yellow line

I got 10 reward 3time and one time - 21reward and then 2time 10reward and then -21reward

why does this happen?

Isn't it cumulative reward means total reward?

Answer 1

[Emmanouil Tzorakoleftherakis]

It's not very clear what you are asking, but the cumulative reward changes during each episode. It also depends on the sample time of the agent (every time the agent takes an action, the cumulative reward gets updated). All things considered, the graph you are showing could make sense (although the values on the y axis are not shown). I would suggest looking at the episode manager plots instead, or you can log and plot the individual rewards throughout an episode