Reliable Robotics and the Certification Path No One Has Walked Before

The Cessna 208 Caravan has been moving cargo for four decades. It is a known quantity — fully certificated, widely operated, deeply understood by the FAA, and the backbone of regional cargo networks for operators like FedEx Feeder and Amazon Air. It is also the aircraft Reliable Robotics has chosen as the platform for its first certified autonomous system.

That choice is not obvious. Building autonomous aviation capability into an existing certificated type is harder to explain to regulators, harder to reason about in a safety analysis, and harder to get right than starting from a clean sheet. It is also, arguably, the faster path to commercial operation at scale — if the certification challenge can be solved.

Understanding why requires understanding what makes Reliable Robotics’ situation genuinely novel.

The Clean-Sheet Path and Why Reliable Robotics Didn’t Take It

Most autonomous aviation programs developing pilotless aircraft are doing so from the ground up. Joby, Archer, Wisk — these companies are certifying new aircraft types under Part 23 or Part 25, working with the FAA to define type-specific certification bases that can incorporate autonomous or single-pilot operation from the start. The regulations can be written around the design. Special conditions can be proposed for novel features. The aircraft doesn’t carry the weight of legacy assumptions.

Reliable Robotics is doing something structurally different. The Caravan already has a type certificate. It was designed for a human crew. Every airworthiness standard it was certificated against — from flight control response to cockpit instrumentation to emergency procedures — was written with the assumption that a qualified human pilot is present, actively monitoring, and capable of intervention.

To install an autonomous flight system on that aircraft and operate it without a pilot onboard, Reliable Robotics must demonstrate that their system meets the intent of those standards, not just the letter. That is a requirements translation problem on a scale the FAA has not formally encountered before.

What “Modification of a Certificated Aircraft” Actually Means Regulatorily

When an applicant modifies a certificated aircraft significantly enough to affect airworthiness, they apply for a Supplemental Type Certificate (STC). STCs are routine for avionics upgrades, engine modifications, and structural changes. They are not routine for removing the pilot from the operational safety chain.

The FAA’s existing framework for STCs assumes the base type certificate remains valid and the modification is bounded in its effects. Autonomous operation of a transport-category aircraft is not a bounded modification in the traditional sense. It touches every aspect of the airworthiness case: flight control, navigation, communication, surveillance, collision avoidance, emergency management, and ground operations. Every one of those domains has existing certification requirements written with human judgment as a variable in the system.

Reliable Robotics’ approach has been to engage the FAA directly on establishing a novel certification basis — effectively negotiating the means of compliance on a requirement-by-requirement basis. This is analogous to what clean-sheet developers do when proposing special conditions, but the structural challenge is different: Reliable Robotics must map autonomous system behaviors onto airworthiness standards that weren’t written to accommodate them, rather than proposing new standards for a new design.

The company has described this process as working with the FAA’s Aircraft Certification Service to define issue papers — formal documents that establish agreed positions on novel certification questions. Each issue paper is, in effect, a requirements specification for how a given airworthiness requirement will be satisfied in the absence of a human pilot.

The System Safety Problem

In conventional aircraft certification, system safety analysis — typically conducted under ARP4761 and documented in a System Safety Assessment — starts from the assumption of a certain crew complement and models failure conditions accordingly. Catastrophic failure conditions are those that would prevent continued safe flight and landing. Hazardous conditions are those that would significantly reduce safety margins or increase crew workload to the point where accurate performance cannot be relied upon.

That last phrase — “crew workload” — appears constantly in advisory circulars, AC 25.1309 guidance material, and FAA policy. It exists because the human pilot is modeled as a mitigating factor in the safety chain. The pilot detects, diagnoses, and responds. System safety analysis for conventional aircraft allocates probability budgets partly on the assumption that some residual failure conditions will be caught by the crew.

Remove the crew, and the entire probability budget has to be re-derived. Reliable Robotics cannot borrow the safety case of the manned Caravan. Every failure mode that previously terminated at “increased crew workload” now terminates at “autonomous system response” — and the reliability, latency, and correctness of that response must be demonstrated to at least the same standard that was previously allocated to the human.

This is not impossible. It is, in fact, an argument that autonomous systems can be made safer than human-piloted ones on specific failure modes where detection latency or cognitive load is the limiting factor. But making that argument to the FAA requires a level of requirements traceability and quantitative safety substantiation that is significantly more demanding than what most manned aircraft programs produce.

Every system safety requirement must be derived from first principles. Every means of compliance must be proposed and agreed. Every claimed failure probability must be supported by test, analysis, or operational data. In domains where operational data doesn’t yet exist — because no one has flown certified autonomous cargo aircraft at scale — the analytical burden is even higher.

What Reliable Robotics Has Actually Built

Reliable Robotics has developed a flight automation system that includes redundant flight computers, datalink architecture, a ground control station, and sensor fusion systems capable of conducting the full flight envelope — taxi, takeoff, cruise, approach, and landing — without onboard crew.

Their ground control architecture is notable because it reframes the human role rather than eliminating it. Operators at ground stations monitor multiple aircraft simultaneously, intervene when necessary, and maintain communication with ATC. This is not fully autonomous in the unmanned-systems sense — it is supervised autonomy, with the human moved from the cockpit to a monitoring role. That framing matters for certification, because it preserves some portion of the human-in-the-loop argument while shifting where and how that loop closes.

The FAA has been specific in its guidance that “remotely piloted” operations face different regulatory treatment than “fully autonomous” operations. Reliable Robotics has been careful to position their system within the remotely piloted framework, at least for initial certification. The ground-based pilot holds a certificate, is type-rated, and is legally responsible for the operation. The autonomous system executes; the human supervises and retains authority to intervene.

That distinction does real work in the certification case. It means the system safety argument does not have to prove that the autonomous system handles every conceivable failure condition without human input — only that it handles the time-critical envelope safely until a ground operator can respond, and that the datalink architecture supporting that response is sufficiently reliable.

The Cargo Aviation Context

Reliable Robotics chose cargo for reasons that go beyond regulatory convenience, though the regulatory advantages are real. Cargo operations under Part 135 already accommodate single-pilot operations on aircraft like the Caravan. There are no passenger manifest complications. Operators are cost-driven and operationally sophisticated. The routes — point-to-point overnight cargo runs, feeder network legs — are often repetitive, low-variability operations where automation delivers reliable value.

The pilot shortage in regional and feeder cargo aviation is genuine and worsening. The supply of ATP-certified pilots willing to fly overnight freight in single-engine turboprops has been tightening for years as the major carriers and cargo integrators absorb the available pool. Autonomous cargo aircraft don’t solve this problem across the entire industry immediately, but they offer operators a path to maintaining or growing capacity without competing for the same constrained labor pool.

The fleet argument is significant. There are approximately 1,000 Caravans operating in cargo service globally. If Reliable Robotics achieves STC approval, every one of those aircraft is potentially addressable. That is not true of any clean-sheet autonomous aircraft program currently in development — each of those creates a new type that must be acquired, trained on, and integrated into maintenance programs from scratch.

What Success Would Actually Mean

An FAA-approved STC for autonomous operation of the Cessna 208 Caravan would be, without overstatement, one of the most significant regulatory milestones in civil aviation since the original certification of Category IIIc autoland systems in the 1970s. It would not just validate Reliable Robotics’ specific system. It would establish a framework.

Issue papers negotiated for the Caravan STC become reference points for the next application. Means of compliance accepted by the FAA for one autonomous cargo aircraft create precedent that other applicants can cite and build on. The safety analysis methodology validated for one system can be adapted for related systems with lower marginal regulatory cost.

This is how FAA certification precedent actually works. The first applicant through any novel certification path carries the highest burden and, if successful, creates value not just for themselves but for the entire subsequent applicant pool.

For cargo aviation specifically, the implications extend beyond the Caravan. A validated certification framework for autonomous operation of certificated turboprops could accelerate programs for the Cessna 408 SkyCourier, the Pilatus PC-12, and other aircraft in the regional cargo fleet. The template is portable even if the STC is not.

The Honest Assessment

Reliable Robotics has been working on this problem since 2017. They have conducted extensive flight testing, engaged seriously with the FAA over multiple years, and built an organization with genuine depth in both flight systems and certification strategy. They are not a concept-stage company.

They are also not done. The certification timeline has extended beyond initial projections, which is normal for novel programs and not a signal of fundamental failure — it is a signal that the FAA is doing its job carefully on a genuinely novel problem. The issue papers that need to be resolved before an STC can be issued are not administrative paperwork. They are substantive technical agreements on hard questions about system safety, datalink reliability, and the appropriate scope of autonomous authority.

The questions are hard because they are the right questions. An autonomous cargo aircraft operating in shared airspace alongside manned aircraft, under instrument flight rules, in degraded weather conditions, is a system where the consequences of a poor safety analysis are not recoverable. The FAA’s caution is appropriate.

What Reliable Robotics has demonstrated is that the path exists and is walkable — that a serious engineering organization with adequate resources and genuine FAA engagement can make meaningful progress on one of the hardest certification problems in contemporary aviation. Whether they complete it first, or whether their work is ultimately completed by a competitor or successor building on their precedent, the requirements framework they are building with the FAA will matter for a long time.

That is not a small thing to have contributed to an industry.