Reinforcement Learning-Based Control of CrazyFlie 2.X Quadrotor

TL;DR

This paper explores combining classical PID control with reinforcement learning to optimize quadrotor control and navigation, including PID tuning and integrating with a positioning system using deep RL methods.

Contribution

It introduces a hybrid control approach that leverages reinforcement learning for PID tuning and navigation of the CrazyFlie 2.X quadrotor, integrating classical and modern techniques.

Findings

Reinforcement learning effectively tunes PID parameters.

Deep Q-Learning enables discrete navigation with predefined motion primitives.

Simulation results demonstrate successful RL-based control and navigation.

Abstract

The objective of the project is to explore synergies between classical control algorithms such as PID and contemporary reinforcement learning algorithms to come up with a pragmatic control mechanism to control the CrazyFlie 2.X quadrotor. The primary objective would be performing PID tuning using reinforcement learning strategies. The secondary objective is to leverage the learnings from the first task to implement control for navigation by integrating with the lighthouse positioning system. Two approaches are considered for navigation, a discrete navigation problem using Deep Q-Learning with finite predefined motion primitives, and deep reinforcement learning for a continuous navigation approach. Simulations for RL training will be performed on gym-pybullet-drones, an open-source gym-based environment for reinforcement learning, and the RL implementations are provided by stable-baselines3