Reliable Robotics and the autonomous cargo aircraft that just flew a Cessna Caravan with nobody on board

Reliable Robotics, a Mountain View startup founded by ex-SpaceX engineers Robert Rose and Juerg Frefel, has flown a Cessna 208 Caravan with no one in the cockpit — no safety pilot, no remote operator on a joystick. The fully automated system handled everything from engine start to parking brake. It is arguably the most advanced autonomous flight system for conventional fixed-wing aircraft in the world, and it signals a fundamental shift in how regional cargo may move within the decade.

What Makes Reliable’s Approach Different?

Most autonomous aviation companies are building entirely new aircraft that happen to be autonomous. Reliable took the opposite path. They started with a certified, proven airframe — the Cessna 208 Caravan, one of the most reliable single-engine turboprops ever built — and added autonomy to it.

This distinction matters for three reasons: certification timeline, economics, and regulatory feasibility. Rather than asking the FAA to certify a brand-new aircraft, Reliable is pursuing a supplemental type certificate (STC), an approved modification to the Caravan’s existing type certificate. That’s a much narrower regulatory ask.

The system itself is not a drone autopilot bolted onto a Caravan. It’s a full redundant avionics suite that replaces the pilot’s decision-making loop. The aircraft carries its own perception system using radar, lidar, and multiple camera arrays, a navigation stack that actively monitors the environment beyond GPS waypoints, and a contingency management system that can divert, hold, go around, or land at the nearest suitable airport — the way a trained pilot would.

What Has Reliable Actually Demonstrated?

In 2023, Reliable completed what they described as the first autonomous flight of a large cargo aircraft in the national airspace system. Not in restricted military airspace. Not over a remote test range. In the NAS, with air traffic control coordination, flying among other aircraft near Hollister, California.

The company also operates a remote operations center where trained operators supervise autonomous flights without directly flying the aircraft. They function more like dispatchers supervising a train than remote pilots. They can issue high-level commands — divert to alternate, for example — but the aircraft executes those commands autonomously. This human-in-the-loop supervisory model is likely what the transition to autonomous aviation looks like for the next two decades.

Why Autonomous Cargo Changes the Economics

The regional cargo network has a cost problem. FedEx and UPS run large jets between major hubs, but the last leg to smaller cities relies on Caravans and King Airs flown by pilots. Pilot salaries, crew scheduling, duty time limitations, hotel costs, and deadhead flights all compress margins on those routes.

An autonomous Caravan that can fly at 2:00 a.m. with no crew duty concerns eliminates most of those costs. It also addresses the pilot supply issue in a segment that struggles to attract applicants. Flying a single-engine turboprop hauling boxes in the middle of the night is not a career destination. If autonomy handles those missions, it frees pilots for work that genuinely benefits from human judgment — complex weather decisions, passenger operations, and training.

What Are the Real Technical Challenges?

Detect and avoid is the biggest hurdle. Manned aviation relies on the see-and-avoid principle. Remove the pilot’s eyes and you need a sensor suite that replicates that capability. Radar and lidar work well for large objects at distance, but detecting a white Cessna against a bright sky at three miles is hard for cameras, radar, and essentially every sensor modality. Reliable’s sensor fusion approach combines multiple detection methods, but the FAA will demand extremely high confidence before approving routine operations in unsegregated airspace.

There’s nuance here, though. The current see-and-avoid system with human pilots is far from flawless. Midair collisions still happen. Near misses happen regularly. The NTSB has said for years that ADS-B and traffic advisory systems are more reliable than eyeballs for traffic separation. The bar autonomy has to clear isn’t perfection — it’s better than the current system.

Weather judgment presents another challenge. Experienced pilots draw on thousands of hours of pattern recognition when evaluating convective activity, icing, and visibility. Teaching a machine to replicate that decision quality using onboard weather radar, datalink weather, and pre-flight analysis is genuinely difficult.

Public trust may be the hardest problem of all. People are uneasy about autonomous cars at 30 mph. An autonomous aircraft at 10,000 feet over populated areas is a harder sell, even if demonstrably safer than a fatigued human pilot. One accident with an autonomous cargo aircraft would set the industry back years, which is why Reliable has been methodical about its testing program.

Who Else Is Working on This?

Xwing (now Daedalus) has also been working on autonomous Cessna 208 Caravans, focusing on retrofitting existing Part 135 cargo operations with extensive testing in the San Francisco Bay Area. Merlin Labs targeted larger aircraft like the King Air 300 and military platforms before being acquired by Palantir in 2025, signaling where the defense community sees this technology heading.

Reliable’s advantages include their team’s SpaceX pedigree — building redundant, safety-critical systems is core to their DNA. SpaceX lands rockets autonomously on drone ships, requiring the same precision, redundancy, and contingency management that autonomous flight demands. Reliable has also stayed focused on a single problem: get the Caravan from A to B without a pilot, safely and reliably, every time.

When Will Autonomous Cargo Caravans Actually Fly Regular Routes?

Reliable is targeting initial commercial operations in the 2026–2027 timeframe, starting with specific routes in partnership with cargo operators. First operations will almost certainly be on routes over sparsely populated areas, probably at night, with significant airspace restrictions — likely rural routes in the western United States where the risk profile is lowest.

The longer-term vision extends well beyond cargo. If an autonomous system can safely operate a Caravan on cargo routes for two years with a safety record matching or exceeding piloted operations, it builds the foundation for larger cargo aircraft, agricultural operations, survey flights, and eventually reduced-crew passenger operations — though that last step is likely decades away.

There’s a useful historical parallel. In the 1920s, airmail routes were the economic engine that proved aviation could be reliable, routine transportation. The technology, infrastructure, and safety culture that emerged from those operations eventually made passenger aviation possible. Autonomous cargo may be this generation’s airmail — the use case that proves the technology, builds the regulatory framework, and creates the safety data for something much bigger.

The Technology That Benefits All Pilots

Even if you never fly alongside an autonomous Caravan, the technology Reliable is developing has direct applications for piloted aircraft. Their perception system can detect runway incursions. Their automated approach system can execute a stabilized approach in zero visibility without the human factors that cause controlled flight into terrain. Their contingency logic handles engine failures with simulator-level consistency rather than the variability of a startled human after six hours in the seat.

Every piece of that technology can make piloted aviation safer.

Key Takeaways

• Reliable Robotics flew a Cessna Caravan fully autonomously in the national airspace system — no pilot, no remote stick operator — using a complete redundant avionics suite with radar, lidar, and camera-based perception.
• Their STC-based certification strategy avoids the complexity of certifying a new aircraft, instead modifying the already-certified Caravan platform.
• Regional cargo economics drive the business case: eliminating crew costs, duty time limits, and scheduling constraints on routes that struggle to attract pilots.
• Detect-and-avoid, weather judgment, and public trust remain the primary technical and social barriers to widespread adoption.
• Commercial operations are targeted for 2026–2027, starting on low-risk rural routes, with a human-in-the-loop supervisory model that blends automation’s consistency with human judgment.