Artificial Intelligence (AI) has become a critical tool for the aerospace industry. It’s helping improve fuel efficiency, part design, inspection processes, and maintenance, repair, and overhaul (MRO). AI is also enabling engineers to give passenger, military, and emergency response craft new capabilities. Some of these can operate fully or semi-autonomously, thanks to AI.

In 2023, aerospace engineers are looking to take AI even higher. At least six aerospace conferences this year plan to focus on intelligent and/or autonomous systems. The American Institute of Aeronautics and Astronautics has called for papers about machine learning to showcase the many ways AI can impact the industry, which is seeing increased demand and expenses after extended COVID-19 lockdowns.

Here's a taste of how AI is taking flight.

AI Unlocks New Potential

In the sky, the big use of AI is for autonomous aircraft. Manufacturers building drones and small planes often spend a great deal of effort on vision systems, and employ AI in sensors and feedback systems that monitor data necessary for a successful flight, including wind, weather, and air traffic.

Kyriakos Vamvoudakis, the Dutton-Ducoffe Endowed Professor at The Daniel Guggenheim School of Aerospace Engineering at Georgia Tech, is currently working on several projects to increase AI’s role in operating flying craft.

“One project involving AI is our development of autonomous land-based vehicles that learn from human-driven vehicles. The point of this work is reinforcement, giving the autonomous vehicles positive rewards if they perform well,” says Vamvoudakis.

The findings can be applied to flying vehicles to reduce the amount of work aerospace engineers need to do to perfect a craft. When the AI agent learns from human behavior, it can make better decisions about a wide range of actions, from handling turbulence to landing after an accident. It can also accomplish these tasks under different conditions, such as a clear day or a snowstorm.

Certain types of aircraft have the potential to be so profitable that several companies are working on them simultaneously. Self-flying air taxis are a popular concept. These planes are sometimes called package or passenger drones, depending on whether they carry objects or people. EHang, a China-based company, Wisk, a California-based company, and Reliable Robotics, which is also based in California and partners with Cessna, are all working on self-flying taxis.

Aurora Flight Sciences is a Manassas, Virginia-based research subsidiary of Boeing that focuses on special purposes unmanned aircraft. One of Aurora’s projects is Centaur, an optionally piloted aircraft system. The system allows a craft to automatically take off and land. It also utilizes an automated contingency management system.

“When operating unmanned in “lost link” mode, the system makes decisions and takes action based on its sensor inputs. The aircraft has a 16-hr flight time and is most often used for reconnaissance or research missions,” says Clint Church, Chief Engineer at Aurora Flight Sciences.

Another Aurora project is MIDAS, an AI-enabled counter-unmanned aircraft system. MIDAS uses deep learning computer vision algorithms to detect, track, and defeat adversarial small unmanned aircraft system (sUAS) targets. The craft is autonomous from launch through landing.

Other new supports for the aerospace industry include hardware-in-the-loop simulators, like Aurora’s Advanced Teaming Integration Lab (ATIL).

“[ATIL] enables user-centric, agile development and [testing] of multi-vehicle autonomous systems. It fills the development gap between pure simulation and flight test, accelerating the design-build-test cycle,” says Church.

Industry leaders rely on partnerships with academic and government institutions to refine AI. Aurora is currently collaborating with MIT on many of its early development programs. One project involves Aurora’s work with MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). Together, the organizations are developing a scalable method to generate and maintain accurate statistical models. They use deep neural networks for object detection and tracking.

Aurora is also working with the Aerospace Controls Lab in MIT’s Department of Aeronautics and Astronautics to develop an AI agent learning methodology. The project optimizes the agents to operate on low size-weight-and-power (SWaP) and communications constraints.

How AI Changes the Role of Engineers

AI is changing the role of aerospace engineers by reducing the need for certain design work. A craft with the right intelligence can find ways to prevent an accident or avoid other problems. AI algorithms that create data on the fly help aircraft chart a course that is more stable and less risky, helping it avoid accidents and problematic weather conditions.

One of Vamvoudakis’ projects uses AI to give drones the ability to develop situational awareness of adversarial activities, such as cyber-physical attacks. A cyber-physical attack is a security breach in cyberspace that could disable a drone. This project involves aircraft learning by providing fast decision making along with guarantees for mission completion and conditions for robustness.

“The guaranteed robustness of the proposed data-driven learning techniques will ensure fast prompt and/or automated decision making without reaching back to higher command levels,” says Vamvoudakis. The project is funded by a $700,000 grant from the Minerva Research Initiative, a U.S. Department of Defense program that supports research in social and behavioral science.

AI can also reduce the turnaround time for servicing jet engines. General Electric says it’s using AI to service its fleet of approximately 70,000 jet engines, and the company has found that so-called digital MRO is better than human intelligence alone at determining when engines need to be taken off an aircraft.

In some cases, AI development does not need to be specific to a given aircraft. Church says Aurora often develops aspects of AI that are vehicle agnostic.

“That means we can port AI-enabled functions from vehicle to vehicle, accelerating the development of new and different aircraft prototypes that are ready to flight test. Additionally, as unpiloted aircraft have become more common, we can leverage autonomy. [Then we] reduce variability and risk when testing unassociated technologies such as propulsion or novel configurations,” says Church.

Aerospace manufacturers make AI-enabled systems in aircraft and spacecraft safer by increasing interactions between people and machines. The term for these efforts is “human-machine teaming.”

“[This] is a key area with many opportunities in advancement, [including allowing] AI to automatically adapt to large changes in vehicle characteristics to help operators retain control. [Human-machine teaming also helps to] estimate and mitigate mode confusion between pilots and the aircraft’s autonomy system,” says Church.

Aurora stresses the need for experienced aviators to be placed in teams with unmanned systems. AI requires good teachers to improve safety, enable new capabilities, and increase trust in autonomous vehicles.

“The physics remains the same. How engineers solve problems and deploy technical capability continues to advance,” says Church.

--

Correction (January 26, 2023): This article originally stated that Aurora Flight Sciences is designing its own self-flying taxis to be overseen by multi-vehicle supervisors on the ground. The company has clarified: "We are no longer designing our own air taxi. Rather, we support our customers who are designing air taxis."