You Can Be My Wingman Anytime

Long-haul flights, which cover vast distances and last for many hours, pose unique challenges to pilots that could potentially endanger the safety and well-being of both passengers and crew. These flights require exceptional focus, endurance, and decision-making skills as pilots navigate a variety of complex situations that can strain their physical and mental abilities. One of the most serious concerns during long-haul flights is the potential for fatigue. Pilots are required to remain alert and attentive for extended periods of time, which can lead to impaired cognitive abilities, slower reaction times, and decreased situational awareness. This fatigue-induced performance decline could have serious consequences, as pilots may struggle to respond effectively to emergencies or rapidly changing conditions.

Emergencies can also put a lot of strain on a pilot's abilities and jeopardize flight safety. These unexpected events, such as engine failures, severe turbulence, or cabin depressurization, necessitate quick and precise decision-making under intense pressure. In these high-stakes situations, pilots must remain calm, communicate effectively with air traffic control, and flawlessly execute emergency procedures to ensure the safety of everyone on board. However, the lengthy duration of long-haul flights can exacerbate the challenges of dealing with emergencies, as pilot fatigue increases the risk of errors and decreased performance during critical moments.

Furthermore, the monotony of long stretches in the cockpit can also lead to complacency, where pilots may become less vigilant in monitoring various flight parameters. This complacency can make them less responsive to subtle changes in the aircraft's behavior or external conditions, increasing the risk of overlooking potentially hazardous situations. Autopilot systems are designed to take some of this strain off of the pilots, and allow them to focus their attention where it is most needed. However, these systems are highly tailored for use in a specific plane, and cannot easily be transferred to another. This makes autopilot systems very expensive and inaccessible for use in most smaller aircraft.

A highly adaptable autopilot system is presently under development by researchers at the Korea Advanced Institute of Science & Technology that can move from plane to plane as easy as a human pilot, but it does not need to go through a lengthy check out process to learn the quirks of each new aircraft like a human does. Something like the inflatable autopilot in the movie Airplane!, this robot takes a humanoid form. Called Pibot, the robot uses computer vision and other sensors to monitor the status of the flight, then leverages mechanical arms and fingers to interact with the unmodified control system of the plane, pressing buttons and adjusting the yoke as needed. Pibot can also verbally communicate with air traffic control during the flight.

Precise control systems allow the robot to accurately manipulate the plane’s controls even during conditions of high turbulence, or during emergencies, which often generate large vibratory forces. External cameras give the Pibot a view of conditions around the plane, while internal cameras give it the information it needs to toggle the controls. Sensor information is fed into artificial intelligence algorithms for processing.

Unlike a human pilot, the robot can memorize virtually endless amounts of navigational data from charts, and it can also be programmed to know the peculiarities of any aircraft. And impressively, with the help of large language models (LLMs), it can glean this information from simply “reading” the preexisting manuals that are written in natural language. LLMs, in conjunction with a speech synthesizer, also enable the Pibot to verbally interact with human air traffic controllers over the radio.

To date, the robot has been tested in flight simulators where it has shown itself capable of error-free flying. In simulated emergencies, it has demonstrated that it can act faster than human pilots to fix problems and take appropriate corrective actions. And of course the Pibot never gets tired or complacent, giving its total focus to the job as long as it is supplied with power.

Pibot can fly solo or act as a copilot from the right seat to give the pilot a bit of relief when needed. But for now, the team is targeting deploying the robot in extreme situations, where it would not be beneficial to have a human pilot in the loop. The robot is expected to be ready to earn its wings in 2026, but the team is still working through some issues before that can happen. For one thing, they need to develop their own LLM that can operate in-plane so that no link to the Internet will be needed while in the air.