Over an alfalfa field in California’s San Joaquin Valley, a small crop-spraying plane is flying scarily low to the ground. There’s little risk to humans, though, as the plane is pilot-free. "We can actually go lower than a human pilot can," says Russ Marotzke, as the aircraft, belonging to the startup Pyka, skims over the crop. Flying lower means less spray drift and therefore less chemicals are needed than in conventional manned crop-dusting, he explains.
Pyka, based in a converted Second World War hangar overlooking San Francisco Bay, designs and manufactures self-flying aircraft without cockpits, intended for either crop spraying or cargo delivery. The company is part of a burgeoning field of startups aiming to bring autonomous fixed-wing aircraft into commercial service. While urban air taxis, known as electric vertical take-off and landing (eVTOL) aircraft, have garnered significant public attention in autonomous aviation, a parallel and quieter race is underway to deploy self-flying planes for more practical applications like crop spraying and cargo delivery. Many of their creators envision these aircraft eventually carrying passengers as well.
"A fully scaled, ubiquitous passenger operation is the holy grail," states Michael Norcia, Pyka’s co-founder and CEO, who envisions a large fleet of Pyka planes, capable of minibus-capacity passenger transport, operating along the US east and west coasts. He believes there’s a "decent chance we’ll get to that point before the eVTOL industry."
This reporter visited one of Pyka’s crop-sprayer test sites, located approximately 80 kilometers (50 miles) east of the company’s factory, accessible via a rough dirt road. On this particular day, Marotzke and a colleague were testing a software update on a demonstration aircraft. Pyka already has about a dozen aircraft operational in Brazil, where they are employed for spraying crops such as cotton and soybeans, tasks previously performed by human pilots. The electric crop-spraying plane features a battery housed in its nose, can fly for approximately 35 minutes, and carries up to 300 liters of spray in a central tank. While sometimes referred to as large drones, Pyka’s aircraft are substantial, boasting an 11.5-meter wingspan.
Within a shipping container situated beside the field, engineers used a computer to designate the target spraying area. The software then autonomously plotted the flight path, factoring in pre-mapped obstacles like nearby power lines. The take-off from a runway adjacent to the field was seamless. Approximately 15 minutes later, the aircraft detected it was running low on spray (water for the demonstration) and executed an autonomous landing for a manual refill and battery swap. It then promptly took to the air again, resuming its precise spraying mission from where it left off.
Autonomous flight fundamentally differs from autopilot. Autopilot systems provide assistance, akin to cruise control and lane-keeping functions in automobiles. Autonomous systems, conversely, are designed to manage the entire flight envelope, including take-off and landing, with minimal to no human intervention, employing complex algorithms to process sensor data and control the aircraft.
The widespread adoption of self-flying planes has lagged behind that of self-driving cars, despite operating in environments generally considered more structured and predictable. Mykel Kochenderfer, an expert in safe aviation autonomy at Stanford University, attributes this partly to major technology companies having "doubled down" on automotive AI, investing vast sums in that sector. Additionally, aircraft are subjected to significantly stricter safety standards than cars, establishing a much higher threshold for deployment. "The consequences for air accidents can just be so severe," Kochenderfer emphasizes.
Military interest has been a significant catalyst for the advancement of this technology. Many companies are engaged in defense contracts, facilitating demonstrations and trials of their systems, often navigating fewer regulatory obstacles than in the civilian sphere. Some are already supplying military customers. In the United States, Pyka’s crop sprayer holds the distinction of being the largest autonomous fixed-wing aircraft approved for commercial civilian use to date, having secured authorization last year. However, these operations are currently confined to a strictly defined agricultural setting and necessitate the presence of a ground operator and a visual observer. Pyka had previously obtained similar approval in Brazil, where regulations are more accommodating. Pyka’s ambitious production goals aim to increase output from approximately two dozen planes annually to 1,000 by 2030, with each unit priced at $550,000 and customers receiving comprehensive training on their operation.
The United Kingdom has yet to grant approval for similar long-term autonomous operations. However, the British firm Windracers is actively seeking permission to launch an autonomous cargo service in the Shetland and Orkney islands. Their aircraft, designed for delivering goods to remote areas, are also currently undertaking missions in Ukraine. Stephen Wright, Windracers’ founder and chairman, asserts, "It would be the first heavy-lift air cargo service by drone certainly in the UK and probably anywhere."
Proponents of autonomous aircraft highlight their potential to alleviate pilot shortages, remove human operators from hazardous tasks like crop spraying, enhance operational efficiency—for instance, by enabling aircraft to carry larger payloads—and reduce costs by allowing a single operator to manage multiple aircraft. They also contend that increased automation could lead to improved flight safety, citing historical reductions in accident rates correlating with the introduction of more automated systems.
However, pilot groups remain apprehensive. The U.S. Air Line Pilots Association (ALPA) has characterized the removal of pilots as "a serious gamble with safety and a step too far." The U.S. National Agricultural Aviation Association, representing crop dusting pilots, expresses concerns that small uncrewed aircraft can be difficult for their aviators to detect, and asserts that piloted planes can cover a significantly larger area more rapidly.
Pyka and Windracers are taking an integrated approach by designing and building their aircraft from the ground up, arguing that this allows autonomy to be a core design principle and the aircraft to be precisely tailored to their intended roles. In contrast, other companies are retrofitting existing larger aircraft with autonomous systems. Reliable Robotics, a U.S.-based company backed by Boeing’s investment arm, is currently testing its system on the Cessna 208B Grand Caravan, a single-pilot cargo plane capable of carrying approximately 1,360 kg of payload over hundreds of kilometers. Robert Rose, Reliable’s co-founder and CEO, explains that retrofitting certified aircraft allows the company to concentrate solely on validating the autonomous system’s safety, rather than simultaneously seeking approval for a new aircraft design.
Merlin Labs, also based in the U.S., has progressively advanced its systems through larger military aircraft and is now applying its technology to the two-pilot Lockheed Martin C-130J military transport plane, with commercial multi-crew cargo planes as the next target. "It is a common autonomy brain that can transition between different aircraft," explains Matt George, Merlin’s founder and CEO.
The companies also exhibit differing philosophies regarding the integration of artificial intelligence (AI). Reliable Robotics is deliberately avoiding AI, believing it would complicate the certification process. Merlin, conversely, is adopting a far more AI-centric approach. This divergence is particularly evident in their development of "detect and avoid" systems. One of the most significant challenges in autonomous flight is replicating a pilot’s ability to perceive and safely maneuver around other aircraft and obstacles, an area where there is virtually no room for error.
With no definitive solution yet established, companies are incorporating various sensor systems and duplicating existing standard sensors to provide additional redundancy. Reliable has integrated forward-looking air-to-air radar to detect other aircraft from over eight kilometers away, coupled with software that adheres to fixed rules to determine the aircraft’s response. Rose confidently states this is "better than a pilot’s eyeballs." Merlin, meanwhile, is utilizing AI-powered cameras to detect and classify objects. Pyka has employed lidar from its inception to identify trees, vehicles, large birds, and terrain. However, due to its limited range, the company plans to incorporate AI-powered cameras, marking its first significant use of onboard AI. "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," remarks Norcia.
The debate surrounding AI extends to communication with air traffic control. In shared airspace, aircraft must be capable of receiving, interpreting, and responding to radio instructions, typically from air traffic controllers. Reliable’s strategy involves a remote pilot on the ground, initially fully trained, to manage communications and make safety-critical decisions. Merlin, on the other hand, intends to employ generative AI, trained on thousands of hours of recorded exchanges, to interpret instructions and respond autonomously. "Our problem is harder…but we want to move beyond remote piloting," states George, who plans to phase out pilots in stages, from two to one, and eventually none. Pyka, according to Norcia, is content to let others "blaze the trail" in determining the most effective methods for operating in shared airspace.
Even if fully autonomous passenger flight remains an aspiration for the future, many anticipate that the technology being pioneered will gradually integrate into commercial aviation, potentially enhancing the safety of piloted flights. As ALPA, the U.S. pilots’ association, notes, such advancements would be a welcome development.







