Multi Agent Reinforcement Learning

I am trying to build multi agent reinforcement learning in 2D grid environment(Combat & Switch) using OpenAI. When I am trying to build game using ten agents, five opponents and the dimension is(15*15) using DDPG algorithm.

So far, I looked into deep reinforce algorithm. However, am not clear why the output of the game is disappearing , such that it allows for modifying entries of the matrix.