An Aircraft Upset Recovery System with Reinforcement Learning

TL;DR

This paper presents an AI-based aircraft upset recovery system using reinforcement learning, specifically the soft-actor critic model, to improve control effectiveness in jet trainers.

Contribution

It introduces a novel RL architecture with hyper-parameter optimization for aircraft recovery, incorporating expert feedback and negative-g punishments.

Findings

AI model's behavior is more desirable than conventional methods

Reinforcement learning with SAC improves recovery control

Incorporates expert feedback and hyper-parameter tuning

Abstract

This article explores the progress made in the creation of a pilot activated recovery system (PARS) for advanced jet trainers that utilizes artificial intelligence (AI) in an effort to enhance operational efficiency. The PARS model employs an advanced reinforcement learning (RL) architecture, incorporating a cutting-edge soft-actor critic (SAC) model and hyper-parameter optimization methods. Negative-g punishments and other handcrafted features remarked upon by control engineers and domain experts regarding PARS are also taken into account by the system. When evaluated by them, the AI model's behavior is deemed more desirable than that of conventional control methods.