Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey on Reinforcement Learning in Aviation Applications

Pouria Razzaghi, Amin Tabrizian, Wei Guo, Shulu Chen, Abenezer Taye, Ellis Thompson, Alexis Bregeon, Ali Baheri, Peng Wei

TL;DR

The survey analyzes reinforcement learning applications in aviation, spanning problem formulations, standard RL methods, and multi-agent extensions, while assessing benchmarks, simulators, and safety/certification considerations. It demonstrates RL as a data-driven framework for planning, control, and maintenance tasks across domains such as collision avoidance, ATFM, revenue management, and flight planning. The paper identifies key gaps in verification, safety guarantees, and sim-to-real transfer, and proposes directions toward runtime assurance, formal verification, and regulatory-ready approaches. By providing a taxonomy, publicly available resources, and a practical roadmap, the work aims to accelerate safe and effective RL deployment in aviation.

Abstract

Compared with model-based control and optimization methods, reinforcement learning (RL) provides a data-driven, learning-based framework to formulate and solve sequential decision-making problems. The RL framework has become promising due to largely improved data availability and computing power in the aviation industry. Many aviation-based applications can be formulated or treated as sequential decision-making problems. Some of them are offline planning problems, while others need to be solved online and are safety-critical. In this survey paper, we first describe standard RL formulations and solutions. Then we survey the landscape of existing RL-based applications in aviation. Finally, we summarize the paper, identify the technical gaps, and suggest future directions of RL research in aviation.

A Survey on Reinforcement Learning in Aviation Applications

TL;DR

The survey analyzes reinforcement learning applications in aviation, spanning problem formulations, standard RL methods, and multi-agent extensions, while assessing benchmarks, simulators, and safety/certification considerations. It demonstrates RL as a data-driven framework for planning, control, and maintenance tasks across domains such as collision avoidance, ATFM, revenue management, and flight planning. The paper identifies key gaps in verification, safety guarantees, and sim-to-real transfer, and proposes directions toward runtime assurance, formal verification, and regulatory-ready approaches. By providing a taxonomy, publicly available resources, and a practical roadmap, the work aims to accelerate safe and effective RL deployment in aviation.

Abstract

Compared with model-based control and optimization methods, reinforcement learning (RL) provides a data-driven, learning-based framework to formulate and solve sequential decision-making problems. The RL framework has become promising due to largely improved data availability and computing power in the aviation industry. Many aviation-based applications can be formulated or treated as sequential decision-making problems. Some of them are offline planning problems, while others need to be solved online and are safety-critical. In this survey paper, we first describe standard RL formulations and solutions. Then we survey the landscape of existing RL-based applications in aviation. Finally, we summarize the paper, identify the technical gaps, and suggest future directions of RL research in aviation.
Paper Structure (22 sections, 3 equations, 3 figures, 6 tables)

This paper contains 22 sections, 3 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Significance and contributions
  3. Preliminaries
  4. Reinforcement Learning in Aviation
  5. Overview of Reinforcement Learning
  6. Standard Formulations of Model-Free Reinforcement Learning
  7. Value-based Methods
  8. Policy-based Methods
  9. Actor-critic Methods
  10. Multi-agent Reinforcement Learning Methods
  11. Selected Applications of RL in Aviation
  12. Public Benchmarks and Datasets
  13. Simulation Environments
  14. Collision Avoidance and Separation Assurance
  15. Air Traffic Flow Management
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Categories of the reinforcement learning models and algorithms.
  • Figure 2: In RL, an agent selects an action $a_t$ based on its current state $s_t$, then it will receive a reward from the environment $r_t$ and arrive to the next state $s_{t+1}$. This process will continue until the agent arrives at a terminal state if any.
  • Figure 3: Taxonomy layout of RL in aviation. Different applications are shown with their corresponding illustrations. Flight planning represents a pre-defined path from the initial point. The revenue management illustration shows an increase in the profit of the airline. Controlling a drone lost one of its motors goes under the adaptive control of an air vehicle. A gimbal shape represents the attitude control of the system. The traffic management sketch depicts the control room monitor to supervise the traffic in the air. A collision avoidance picture is an alarm of avoiding a conflict between two vehicles.