You can thank the aerospace industry for everything from Velcro to anti-lock brakes. When new technology needs to be developed, it is hard to go wrong entrusting engineers who are, quite literally, doing rocket science on a daily basis. Could a partnership between aerospace and automotive specialists be beneficial for both?
- Aerospace and automotive autonomous systems have the same goal, but are approaching it from different ends of the problem.
- The aerospace industry has decades of experience of using autopilot systems, but only in the air.
- Airbus is working on a vision-based autopilot, similar to what you find in autonomous cars.
As complex as modern cockpits look, all that technology has made it far easier to pilots to do their jobs. (Photo: Getty Images)
History of autopilot
The promise of autonomous cars has been around since the 1950s, but 70 years later, we are still struggling to make it happen. Yet Autopilot has existed in the aerospace industry for over a hundred years. The Sperry Corporation developed a working autopilot system for aircraft in 1912. It used two gyroscopes, one for up and down and the other for left and right. It kept the plane flying straight and level at a preselected heading and altitude.
In the 1930s, Radio Direction Finding or RDF systems used transmitters on the ground and receivers in airplanes. It allowed planes to home-in on a particular signal to find their bearing. Although RDF signals are still in use, the same GPS in cars and cell phones is the best method for automated navigation in planes, at least to get to a general location in the air.
GPS can navigate a plane on a preprogrammed route, but it isn’t accurate enough to actually land a plane. Large commercial airliners use Instrumented Landing Systems or ILS. ILS uses focused radio signals to essentially funnel the airplane down onto the runway. Even with all that tech, most landings are still flown by the human pilots.
So autopilot is great at getting to a destination, but it’s missing the fine control necessary to land and any sort of system to keep planes from hitting things once on the ground. But that could be changing. Airbus, Europe’s leading aerospace company, is currently developing a vision-based autonomous taxiing and take-off system for airplanes maneuvering on the ground at airports.
Vision based autonomous systems are essential for following lines on taxiways and runways. (Photo: Getty Images)
Autopilot vs. autonomous cars
Autonomous driving systems in cars, and autopilot in planes are hoping to accomplish the same goal, but they are approaching it from opposite directions. Airplanes are evolving from a system that once the plane was pointed in a particular direction, it stayed straight and level. Cars, on the other hand, aren’t heading to a destination; from the beginning of autonomous driving, the goal has been to follow the road – visually.
A self-driving car and a self-flying airplane have very different requirements, partly because aircraft autopilot systems are focused on flying the plane in a big wide-open sky, not taxiing around the ground. Current semi-autonomous cars are loaded with other sensor to monitor for obstacles, but ultimately, if you want a car to follow the road, it has to be able to see it. It has only been in the last couple of years that the aerospace industry has really taken vision-based systems seriously.
The goal of autopilot for aircraft was originally to allow the pilots to do more important tasks than simply keeping the plane level. Up until the 1950s, airliners had 5-person crews to handle everything from navigation, to communications, to keeping track of the plane mechanically. Now, most have 2-person crews. Airlines would now like to drop that to one.
On top of cost savings, it is believed that autopilots will have fewer errors. This is also one of the main arguments for autonomous cars. The vast majority of car accidents are caused by human error. A study from the Eno Center for Transportation suggests over 20,000 lives per year could be saved if 90% of cars in the United States were autonomous.
Airport taxiways and tarmac may look like complete chaos, but they are strictly controlled, a perfect environment for gathering the machine learning data required to train autonomous systems. (Photo: Getty Images)
What can aerospace offer the auto industry?
We’ve all heard that flying is safer than driving and one of the reasons for that is the aerospace industry’s absolute adherence to redundant systems; always have a backup to your backup. Current semi-autonomous automotive systems are rendered inoperable if even one sensor is compromised. Anything from a dirty camera lens to poor weather can affect a car’s self-driving ability. At 30,000 feet, pulling over and waiting for a tow truck isn’t an option, so aerospace engineers always think three solutions ahead of a problem. So first, aerospace could bring the redundancy engineering that doesn’t exist in the automotive space. Eventually, autonomous cars won’t have steering wheels, so driver intervention won’t be an option.
Another thing autonomous systems need to succeed is, quite simply, time. Machine learning is the process of computers “figuring out” how to function in different situations. Airplanes will log thousands of hours of operation in semi-controlled environments while being supervised by trained pilots. Sounds like a better alternative than Teslas being trained by Instagram-influencers. Much like machine learning, most of human-machine-interface research is extremely time intensive. Designers can conduct studies and simulations, but it doesn’t replace real data from actually being used in the environment.
This begs the question, what can the automotive industry offer aerospace? Currently, much of what aerospace is doing revolves around camera-based systems. But if planes are going to use autopilot for taxiing around airports in all weather conditions, day or night, they will need to adopt radar or lidar systems. Car companies have several years of experience under their belts with these and could essentially offer aerospace turn-key systems.
WHY THIS MATTERS
The auto industry has been over-promising on the timeline of autonomous cars for a couple of decades. Whenever a manufacturer makes major progress on a self-driving car, they learn the problems in the real world are harder than expected. Thousands of hours of machine learning, and huge amounts of engineering knowledge that will be made available through the inclusion of aerospace innovators could speed up the development by an order of magnitude.