RL in dynamic environment

I am trying to apply RL in marine environment. An example is when a fish swims toward its altimate goal, it should avoid bad weather or predator areas. We can call these areas as obsticals. However these obesticals are not fix, they change over time. That is, if the fish today is in cell(3,3), and cell(3,4) is the only safe nbr cell, but tomorrow it could be not safe anymore. I managed to implement the Q-RL but for one static environment ( all obstical areas are fixed dueing the whole simuation). Would you please tell me if it is possible to appy for such dynamic env. If you would suggest some resources/ref I would be grateful.

Answer 1

[Emmanouil Tzorakoleftherakis]

The following example seems relevant, please take a look:

https://www.mathworks.com/help/robotics/ug/avoid-obstacles-using-reinforcement-learning-for-mobile-robots.html