Eidahn Eliash

Date of Award

3-2023

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Aeronautics and Astronautics

First Advisor

Donald L. Kunz, PhD

Abstract

Recent research has been conducted in using deep reinforcement learning algorithms to generate closed form optimal control laws for helicopters in power loss scenarios. The control laws could allow for design of various cockpit visual pilot aids if proven successful. Such pilot aids are predicted to reduce pilot workload during execution of an emergency powerless landing, and to overall increase the probability of landing safely in such an event. The proposed topic discusses the use of the PPO algorithm in order to find an optimal control policy for a helicopter in a total power loss emergency landing scenario. The task consists of designing and training an actor neural network and a critic neural network using the PPO algorithm, along with developing the dynamic equations of motion for the helicopter, defining a reward structure and tweaking relevant hyper-parameters. The PPO algorithm has been proven overall feasible for bringing a simplistically modeled helicopter to safe landings over a wide range of initial flight conditions.

AFIT Designator

AFIT-ENY-MS-23-M-267

Comments

A 12-month embargo was observed.

Approved for public release. Case number on file.

Download

Share

COinS