Reinforcement Learning Based Autopilot
Establishing autonomous flights will have big consequences for the future and the way the aviation industry works. To that end, will explore the feasibility of RL algorithms in autonomous flight by breaking this gargantual goal into smaller ones.
The goal of this project is to be able to train Autopilot RL agents using
Policy Gradient algorithms. While our goal is to create an Agent that is capable of various flight maneuvers,
for the course of this project we will be working on implementing an agent capable of changing its altitude as desired.
During the course of this project, we will explore different RL strategies in order to determine the most suitable approach
for our application.
We have discussed the implementation of all the agents and their performances in our project documentation below.
The agents that we have implemented are:
- REINFORCE
- Proximal Policy Optimization (PPO)
- Deep Deterministc Policy Gradient (DDPG)
- Soft Actor-Critic (SAC)
