I'm working on training a game AI using deep reinforcement learning to achieve specific examples based on pixel input and some additional state information.
Naturally, I'm using a convolutional neural network to deal with the pixel information, which has been working well so far. However, I still have additional information available, such as numeric values associated with current health and ammo.
I know it must be possible to create a network architecture to take advantage of both the spatial information provided by the screen buffer as well as non-spatial information such as the game states I mentioned earlier. Is there any way to create a neural net architecture that can handle both spatial and non-spatial state?