An arXiv preprint reports a single-stick control concept designed to guide a transformational eVTOL through hover, transition and cruise without direct mode switching. In a representative simulation, the aircraft model remained stable across those phases and produced smooth, low-jerk trajectories.
The reported evidence comes from representative mission simulations, motion-simulator observations and model-based optimal-control trajectories rather than a real-aircraft validation. The paper describes its human-factors evidence as preliminary and primarily qualitative.
One interface across changing flight regimes
The proposed interface keeps one inceptor—the pilot’s main control stick—throughout the flight envelope. It is designed to preserve the meaning of the pilot’s inputs as the aircraft changes from hover to transition and then to cruise, avoiding direct mode switching. An active side stick provides force feedback to guide the pilot.
The control architecture combines three elements: unified command filtering, full-envelope inversion-based control and active side-stick force feedback. In ordinary terms, the system filters pilot commands before they reach the aircraft, uses a control law intended to work across the flight envelope, and feeds forces back through the stick.
The design seeks to keep the pilot’s input interpretation consistent as the aircraft moves between flight phases. In the reported representative mission, the command-filtered trajectories were smooth and low in jerk, while the aircraft operated stably through hover, transition and cruise.
Early simulator observations remain limited
The pilot-in-the-loop work used the DLR PAVSIM two-seat cockpit, mounted on an industrial robot and equipped with a force-feedback side stick. The text reports one recorded flight. It does not report the total number of participants, the number of trials or a demographic description.
The reported pilot feedback was qualitative rather than a statistical measurement of workload or handling quality. The authors say non-expert participants adapted quickly, while conventionally trained pilots needed additional familiarization. Little stick activity was observed during the generic mission.
Those observations cannot establish that the system reduces pilot workload, improves safety or is easier to learn across pilots. The study does not report a larger, statistically powered human evaluation or quantitative workload results.
The smoother option took longer in the model
The researchers also used optimal control, a method that searches for a feasible aircraft trajectory under specified constraints, to compare five cases within one mission scenario. The calculation used direct collocation with trapezoidal integration on an evenly spaced grid of 60 nodes per flight phase.
The reported final time was 113.3 seconds for case 1. Each of the command-filter cases 2a, 2b and 2c took 127 seconds, while the open-loop case took 58.1 seconds in the stated model-based comparison.
The paper reports that the control-inceptor mission was around 12% slower than the closed-loop aircraft case. That comparison describes a model-based scenario, not a direct measure of how quickly a pilot could complete the mission. The optimized trajectories represent best-case solutions under the model and constraints used in the calculation.
The force-feedback comparison also needs careful reading. In the two input-penalized cases, 2b and 2c, the final time was 127 seconds in both. Their combined objective values were 281.8 and 261.7, respectively. Because that objective combines mission time with a penalty on inputs, the difference is not a standalone measure of reduced stick activity.
What the tests leave unanswered
The simulation did not include sensor latency, wind and turbulence, failures or other off-nominal conditions.
The simulation evidence therefore does not show performance in a real aircraft, and the optimized trajectories are not direct measures of pilot performance.
The authors say stronger human-factors conclusions require a larger, statistically powered study. Further testing would also need to examine the conditions omitted from the simulation, including latency, wind, turbulence and failures.
The manuscript is an arXiv preprint, version 1, dated 20 August 2026. The authors state that no funds, grants or other support were received during preparation and report no conflict of interest. They say the article or cited methods contain the data and methods needed to replicate the approach, with no extra data required.
Paper data and sources
Original title: Taming the Tilt: A Unified Pilot Control Concept for Transformational eVTOL Aircraft
Authors: Daniel Milz, Marc May, Andreas Seefried, Tobias Bellmann
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text