Adding new inputs after a convolutional layer in a neural network?

Question

I'm doing Reinforcement Learning to train autonomous navigation for a drone. The supported problem is camera based autonomous navigation thanks to a simulated environment and a CNN: AlexNet. I want to add two other inputs to optimize the navigation: the position/orientation of the drone and the volume unit traveled in the form of a 3D matrix. The purpose of the neural network is to predict the next action of the drone.

I want to add those two new inputs after the convolutional layers but I don't know how to integrate them. I intent to flatten the 3D matrix and to add it along with the position/orientation before the fully connected layer but I don't know if it's the right way to do it.

What kind of network is the best to integrate those two new inputs?

Answer 1

You might want to inject the new inputs by concatenating them to a layer after flattening.

I suggest you check the stackoverflow link: How to concatenate two layers in keras?

for an example and good explanation