Deploying the Agent

Deploying an agent means exporting the agent, loading it into a control system, and then asking the agent for decisions in a real destination environment. The export will be a combination of Python files and model (.mdl) files that contain the neural network weights used by skills trained with DRL.

For example, if you build an agent to pilot a drone, after you export the agent you will place the resulting files onto the computing processor of the drone. Then you will write code that passes drone sensor information to the agent, asks it for the next action, and then passes the action from the agent to the rotors of the drone.