Aerofugia Study: Pilots' eVTOL Safety Role
The study presents human involvement as part of a layered safety architecture under current technological and operational constraints.
Researchers at Aerofugia have published a paper titled "The Necessity of a Human Pilot in eVTOL—Balancing Safety and Autonomy" in Aerospace (Volume 13, Issue 5, Article 412; DOI: 10.3390/aerospace13050412), published on April 28.
And it’s available to download for free.
The study examines the role of human pilots in eVTOL operations, with a focus on safety in low-altitude urban airspace.
It adopts a system-level framework combining probabilistic simulation, statistical analysis of CAAC general aviation incident data (2017–2021)—including bird strikes, encounters with high-altitude obstacles and airborne objects (with unauthorized or uncontrolled UAV activity accounting for approximately 61% of such incidents, or 57 cases during 2017–2021, representing a major contributing subset)—and fault tree modelling to assess operational risks under both autonomous and human-assisted configurations.
The safety assessment is referenced against established civil aviation requirements, including a target catastrophic failure probability of less than 10⁻⁸ per flight hour.
To characterize external hazards, the authors develop a Monte Carlo simulation framework modeling random encounters between an eVTOL aircraft and non-cooperative UAVs.
The scenario represents a low-altitude descent corridor (1,000 m × 150 m, altitude band 50–150 m), with the eVTOL following a nominal 3° glide slope at 20 m/s and UAV motion modeled stochastically at approximately 10 m/s. A conservative collision threshold of 10 m separation is applied.
Across 10,000 simulation runs, the estimated collision probability is approximately 0.18%. The analysis is explicitly intended as an order-of-magnitude estimate rather than a high-fidelity physical prediction.
While direct collision likelihood is low, a substantial proportion of encounters fall within the 100–200 m separation range, representing a critical interaction zone where continuous monitoring, prediction, and conflict resolution remain necessary.
System-level safety is further evaluated using fault tree analysis applied to a representative eVTOL platform, the AE200, described as a tiltrotor aircraft with distributed electric propulsion. The analysis considers a failure scenario involving a single motor failure combined with loss of autonomous fault detection and control response.
Under nominal assumptions, the resulting system-level failure probability can be consistent with the 10⁻⁸ per flight hour target. However, the authors also introduce a more conservative and operationally realistic scenario in which the combined failure probability is adjusted to approximately 10⁻⁶, reflecting increased uncertainty in dense, low-altitude urban airspace.
In the aforementioned degraded case, human pilot intervention is modelled as an additional safety layer.
Using a conservative human error probability on the order of 10⁻²—decomposed into omission and incorrect action events consistent with established human reliability assessment methods—the analysis shows that pilot involvement acts as a compensatory mechanism, enabling the system to maintain the target catastrophic failure probability under degraded conditions.
The paper also introduces a scenario-based operational framework distinguishing between routine and off-nominal conditions, with varying levels of automation and pilot involvement across flight phases. Minor deviations are handled autonomously, while more significant or time-critical situations require pilot monitoring or intervention.
Moreover, the study outlines a near- to medium-term deployment pathway centered on hybrid human–machine operation.
It discusses associated engineering, economic, and regulatory considerations, including sensing system size, weight, and power constraints; infrastructure and cybersecurity requirements for highly connected operations; and compatibility with existing certification frameworks such as FAA powered-lift regulations and EASA Special Condition VTOL, which currently assume the presence of a pilot-in-command.
The authors emphasize that the analysis is an engineering-level approximation rather than a certification-grade validation, with simplified models used for tractability.
They also highlight limitations of current autonomous systems, including challenges related to environmental uncertainty, non-deterministic behavior of AI-based decision-making, and the difficulty of demonstrating deterministic, verifiable performance under certification standards.
Overall, the study presents human involvement as part of a layered safety architecture under current technological and operational constraints.
It supports a phased transition toward higher levels of autonomy, with hybrid human–machine collaboration serving as a practical and scalable intermediate step as system capabilities, infrastructure, and regulatory frameworks mature.