The technology is being tested first in areas where the risks and regulatory demands are more manageable, including agricultural spraying and cargo delivery. But companies developing the systems have a much bigger ambition: to eventually put passengers aboard self-flying aircraft.
At the forefront of that effort is Pyka, a California-based start-up developing autonomous fixed-wing aircraft without cockpits. The company's planes are designed primarily for crop spraying and cargo transportation, with passenger operations representing a longer-term goal.
Over an alfalfa field in California's San Joaquin Valley, one of Pyka's aircraft demonstrates just how different autonomous aviation could look. The small spraying plane flies remarkably close to the ground as it moves across the crop, but there is no pilot sitting inside.
"We can actually go lower than a human pilot can," says Russ Marotzke, a Pyka flight test engineer.
The lower flight path is not simply a technological demonstration. According to Marotzke, flying closer to the crop reduces spray drift, meaning less chemical can be required to cover the same area compared with conventional manned crop-dusting operations.
The aircraft is fully electric, with its battery positioned in the nose and a spray tank located in the centre of the fuselage. It can remain airborne for about 35 minutes and carry as much as 300 litres of liquid.
Although Pyka's aircraft are sometimes described as large drones, their scale is considerable. Each has a wingspan of about 11.5 metres.
At a test site roughly 80km east of Pyka's factory, engineers recently used one of the aircraft to test a software update. The location, reached by a rough dirt road, provides a controlled environment in which the company can demonstrate how the system manages an agricultural mission without an onboard pilot.
From a shipping container beside the field, engineers use a computer to identify the precise area to be treated. The aircraft's software then calculates a flight route, incorporating information about previously mapped hazards such as power lines.
The take-off is largely uneventful. The aircraft accelerates along a runway beside the field, rises into the air and begins its programmed mission.
Around 15 minutes later, after determining that its supply of water — used instead of chemicals for the demonstration — is running low, the aircraft returns and lands itself. Engineers refill the tank and demonstrate a battery change before sending it back into the air.
The plane then resumes its work at precisely the point where it stopped.
That distinction is important because autonomous flight is not simply another name for autopilot.
An autopilot assists a human pilot in controlling an aircraft, much as cruise control and lane-keeping systems assist a driver. Fully autonomous systems, by contrast, are intended to manage the entire flight, including take-off and landing, by processing information from sensors and making decisions about how to control the aircraft.
Despite the apparent parallels with self-driving cars, autonomous aircraft have taken longer to emerge.
Mykel Kochenderfer, a Stanford University expert in aviation autonomy and safety, says part of the reason is that major technology companies invested heavily in autonomous cars, effectively "doubled down" on the sector and poured enormous amounts of money into its development.
Aviation also faces a much higher safety threshold.
"The consequences for air accidents can just be so severe," says Kochenderfer.
For companies seeking approval to operate autonomous aircraft, demonstrating that their systems can cope with unexpected events is therefore one of the central challenges.
Military demand has nevertheless helped accelerate development. Several companies working on autonomous aviation have defence contracts to test and demonstrate their systems, where regulatory requirements can differ from those governing civilian commercial operations. Some developers are already supplying military customers.
Civilian aviation is beginning to follow.
In the United States, Pyka's crop-spraying aircraft is currently the largest autonomous fixed-wing aircraft approved for commercial civilian operations. The company received authorisation last year, although the approval applies only to tightly defined agricultural operations and still requires a ground operator and visual observer.
Pyka had previously secured similar approval in Brazil, where the regulatory environment is more permissive.
The company says about a dozen of its aircraft are already operating in Brazil, spraying crops including cotton and soybeans — work traditionally performed by human pilots.
Pyka now wants to dramatically increase production. From roughly two dozen aircraft a year today, it hopes to reach a manufacturing rate of 1,000 planes annually by 2030. Each aircraft sells for about $550,000, with customers receiving training to operate the systems.
The United Kingdom is also beginning to explore the technology.
British company Windracers is seeking approval for an autonomous cargo operation serving the remote Shetland and Orkney islands. Its aircraft are intended to transport goods to areas where conventional logistics can be difficult and expensive.
The company's aircraft have also been used for missions in Ukraine.
"It would be the first heavy-lift air cargo service by drone certainly in the UK and probably anywhere," says Stephen Wright, Windracers' founder and chairman.
For supporters, the potential benefits extend well beyond eliminating the need for a person to sit in a cockpit.
Autonomous aircraft could help address pilot shortages, remove people from hazardous jobs such as agricultural spraying and increase operational efficiency. Cargo aircraft, for example, could potentially be designed around the available payload rather than the needs of a human crew.
There is also the prospect of reducing operating costs if a single person on the ground can supervise several aircraft.
Developers argue that automation could eventually make aviation safer, pointing to the industry's historical reduction in accidents as increasingly sophisticated automated systems have been introduced.
Pilot organisations, however, remain deeply sceptical about removing human pilots.
The US Air Line Pilots Association (ALPA) has described the idea of eliminating pilots as "a serious gamble with safety and a step too far".
Agricultural aviation groups have raised their own concerns. The US National Agricultural Aviation Association says small uncrewed aircraft can be difficult for pilots to see and argues that conventional manned aircraft can spray significantly larger areas in less time.
The technological race is also producing competing approaches.
Pyka and Windracers are designing aircraft specifically around autonomous operation. Their argument is that building an aircraft from scratch allows autonomy to be incorporated into the design from the beginning.
Other companies are taking the opposite approach by adding autonomous systems to aircraft that are already certified and in service.
Reliable Robotics, a US company backed by Boeing's investment arm, is testing autonomous technology on the Cessna 208B Grand Caravan, a single-pilot cargo aircraft capable of carrying roughly 1,360kg over hundreds of kilometres.
Robert Rose, the company's co-founder and chief executive, says using an already certified aircraft allows Reliable Robotics to concentrate on proving the safety of its autonomous system rather than simultaneously certifying an entirely new aircraft.
Merlin Labs, another US company, has taken a different route. It has tested its systems on progressively larger military aircraft and is now working with the two-pilot Lockheed Martin C-130J transport aircraft, with commercial multi-crew cargo aircraft planned as a subsequent step.
"It is a common autonomy brain that can transition between different aircraft," explains Matt George, Merlin's founder and CEO.
One of the biggest divisions between the companies concerns artificial intelligence.
Reliable Robotics is deliberately avoiding AI in its core autonomous flight system, arguing that using AI could make the technology more difficult to certify.
Merlin, by contrast, is taking a more AI-focused approach.
The difference is particularly clear when it comes to "detect and avoid" technology — systems designed to replicate one of the most basic but critical tasks performed by pilots: identifying other aircraft, obstacles and potential hazards and responding before a collision occurs.
There is virtually no room for error.
Companies are therefore experimenting with combinations of radar, cameras, lidar and other sensors, while also duplicating critical systems to provide additional protection if one component fails.
Reliable Robotics has installed forward-looking air-to-air radar capable of detecting other aircraft more than eight kilometres away. Its software then uses predetermined rules to determine how the aircraft should respond.
"It is better than a pilot's eyeballs" says Rose.
Merlin is pursuing a different strategy, using AI-powered cameras to identify and classify objects.
Pyka has relied on lidar from the beginning to detect hazards including trees, vehicles, large birds and terrain. Because lidar has a limited range, however, the company plans to supplement it with AI-powered cameras.
That will represent Pyka's first significant use of AI directly onboard its aircraft.
"For a lot of things there's no need to use AI…but for figuring out that six pixels in the distance are an airplane versus some other smudge, it is perfect territory," says Norcia.
Artificial intelligence could also change how autonomous aircraft interact with air traffic controllers.
Aircraft operating in shared airspace must be capable of receiving, understanding and responding to radio instructions. Today, Reliable Robotics intends to keep a remote pilot on the ground, initially with that person fully trained to communicate with air traffic control and make critical safety decisions.
Merlin wants to go further. The company plans to use generative AI trained on thousands of hours of recorded communications to interpret air traffic instructions and eventually respond without a human pilot.
"Our problem is harder… but we want to move beyond remote piloting," says George.
Merlin plans to reduce the human role gradually — moving from two pilots to one and eventually to none.
Pyka's Norcia is taking a more cautious position on the question of autonomous aircraft operating in complex shared airspace.
The company, he says, is content to let other developers "blaze the trail" as the industry determines how best to manage autonomous aircraft alongside conventional manned aviation.
For now, the passenger aircraft envisioned by Pyka and others remain some distance away.
Norcia nevertheless sees passenger transport as the ultimate prize. He imagines fleets of Pyka aircraft with enough capacity to function like airborne minibuses, carrying travellers between cities along the US east and west coasts.
"A fully scaled, ubiquitous passenger operation is the holy grail," says Michael Norcia, Pyka's co-founder and CEO.
"There's a decent chance we'll get to that point before the eVTOL industry."
Electric vertical take-off and landing aircraft, or eVTOLs, have attracted much of the public attention surrounding the future of autonomous aviation. Yet the development of self-flying conventional aircraft could prove just as significant.
Rather than attempting to revolutionise passenger travel immediately, autonomous fixed-wing aircraft are entering aviation through more practical applications — spraying crops, transporting cargo and undertaking missions in areas where putting a human pilot at risk may be undesirable.
Even if fully autonomous passenger flight takes many more years to become routine, the technology developed along the way could gradually find its way into conventional aviation.
It could improve navigation, collision avoidance, communications and other safety systems, potentially assisting rather than replacing pilots.
For ALPA, that would be a far more welcome direction for the technology.
The future of aviation, therefore, may not arrive as a sudden moment when pilots disappear from cockpits. It may emerge incrementally, beginning in farms and freight routes before spreading into increasingly complex areas of commercial flight.
The aircraft may be pilotless. The race to make them safe, however, is being driven by very human concerns.