Deep Learning and Reinforcement Learning: Embedding Artificial Intelligence into a Flapping-Wing MAV
This repository hosts the ROS Workspace for a closed loop control system that uses Deep RL (DDPG) to teach a real physical flapping wing MAV to fly. This workspace was created in ROS Melodic (Ubuntu 18.04) The physical MAV looks as such
The closed loop hardware and software cycle is designed as such
The information in the integrated framework phase flows through ROS nodes and topics in the following manner
All details about the setup and training process can be found in FWMAV_DeepRL_Final.pdf