I am new to Reinforcement Learning. I am trying to train PPO agent for citylearn. The goal is to lower two environmental variables from observations. The default reward function is

reward = (variable1 + variable2)*-1

The agent is providing actions with values that are close to -1 for higher iterations. Instead of attempting to improve the environment, it only tries to avoid high negative rewards. I've tried normalizing observations and rewards, but it's still been ineffective.

Is there a reward function that would encourage exploration rather than the same behavior?


1 Answer 1


There is no one-size-fits-all answer to this question, as the best reward function for encouraging exploration will vary depending on the specifics of the AI system. However, some possible reward functions that could encourage exploration include giving the AI a small reward for each new state it visits, or giving the AI a large reward for discovering new states that are particularly valuable.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .