Hi everyone!! I’m training an agent in the Lunar Lander environment. I trained the agent with approximately 800 episodes. I ran tests on other episodes, so I trained the agent for more episodes and decreased the epilon decay value.
However, in some test episodes, the agent doesn’t fully land; it just ‘floats’ over the surface, but doesn’t land, and goes upwards.
Why would this be? I trained it for more epochs and decreased the epsilon decay.
Thanks! 
