One Stick, All Flight Phases: Taming the Trickiest Part of Flying an Air Taxi

Imagine stepping into a cockpit where the fifteen-odd levers, pedals, and switches that once demanded a pilot's undivided attention have been replaced by a single stick — and where that stick gently presses back against your hand, physically guiding you through the trickiest part of the flight. That is the promise of the control concept at the heart of this paper by researchers at the German Aerospace Center (DLR), who set out to answer a deceptively simple question: what does flying an electric vertical take-off and landing (eVTOL) aircraft need to feel like so that a person who isn't a professional test pilot can actually do it?

The Science

Electric vertical take-off and landing aircraft — the small, battery-powered air taxis and delivery drones that have dominated aviation headlines for a decade — are easy to conceptualize and brutally hard to fly. They operate in three distinct regimes that demand almost opposite instincts. In hover, they behave like helicopters: the pilot thinks in terms of drifting sideways, climbing, descending, and spinning about a point. In cruise, they behave like fixed-wing airplanes: the pilot thinks in terms of airspeed, flight path angle, and coordinated turns. And in the transition between the two — where the aircraft's wings tilt from vertical to horizontal — neither instinct works, and the machine passes through its most fragile and unstable territory.

Historically, this transition was a pilot's nightmare. Early rotorcraft prototypes like the Bell XV-3 and XV-15 demanded manual coordination of control surfaces, engine power, and pylon tilt angles, a cognitive load that contributed directly to high accident rates (Maisel et al., 2000). The Harrier jump jet's vertical landing was famously called the "three-hand problem," because pilots had to operate the thrust lever, the stick, and a separate nozzle-angle lever simultaneously (Nordeen, 2006). The military solved this with digital fly-by-wire systems — the F-35B, for instance, uses a unified control scheme with an active side stick that physically moves and pushes back against the pilot (Wurth and Smith, 2018). But those solutions were designed for elite fighter pilots with hundreds of hours of training.

The civilian eVTOL industry faces a different constraint. If these aircraft are ever to become ubiquitous — ferrying passengers across cities, delivering cargo, serving as personal transport — they will need to be flown by people who are not professional aviators. NASA has formalized this ambition as Simplified Vehicle Operations (SVO): the goal of enabling safer operation by less-experienced pilots through lower workload and reduced training demands (Wing et al., 2020).

This is the context for the DLR team's work. The researchers — Daniel Milz, Marc May, Andreas Seefried, and Tobias Bellmann, all at DLR's Institute of Flight Systems in Weßling, Germany — built a control concept around a specific aircraft: a tandem tilt-wing eVTOL with eight electric propellers, two independently tilting wings, and four elevons, giving it 14 separate control inputs. They then set that aircraft's flight dynamics into a full-motion simulator equipped with an active, three-axis force-feedback side stick — a joystick that does not just read the pilot's inputs but pushes back, generating tactile cues. Their aim was to design a single, unified control interface that works identically across hover, transition, and cruise, with no mode switching, no reinterpreting of what a stick deflection means, and no need for the pilot to think about the aircraft's physics at all (Milz et al., 2026).

The technical backbone is a control architecture called hybrid nonlinear dynamic inversion, combined with optimization-based control allocation. In plain terms, the flight control system continuously inverts the aircraft's dynamics — it works backward from "the pilot wants to move this way" to "here is exactly how much thrust each propeller must produce and how far each wing must tilt." This is not new in itself; similar approaches exist (Raab et al., 2018; Lombaerts et al., 2020a). What distinguishes this work is the integrated combination: unified command filtering, full-envelope inversion-based control, and an active side stick with force feedback, all working in one operational framework, validated on a full-motion simulator.

Unified control is a specific technical term here, and it matters. It means the command semantics are consistent across every flight phase. Whether the aircraft is hovering at a standstill or cruising at 70 meters per second, a given stick deflection means the same thing to the pilot. The inner-loop flight control handles the configuration-dependent actuation automatically — the pilot never selects a "hover mode" or a "cruise mode." The concept satisfies five design requirements drawn from EASA's draft certification rules for VTOL aircraft (SC-VTOL-01): easy to operate and learn; releasing the stick at any time continues the current flight state safely; unified control throughout the whole envelope; transparent inceptor effect so the pilot always knows what an input will do; and minimal performance penalty from the command systems.

The heart of the implementation is a set of command filters that translate raw stick deflections into the controlled-variable commands the aircraft tracks. These filters blend several established techniques. Hover damping provides intuitive velocity control in slow flight. The Height Above Terrain (HAT) funnel shapes how the aircraft manages altitude near the ground. Translational Rate Command (TRC) lets the pilot directly command a horizontal velocity rather than a bank angle (Lombaerts et al., 2020b; Grondman et al., 2018). The filters are tuned against standard rotorcraft and fixed-wing handling qualities specifications, chiefly ADS-33 (Samuel Crews et al., 2000), and impose the aircraft's response on the closed loop through a second-order reference model.

Crucially, the filters are continuous — they blend smoothly from hover behavior to cruise behavior, rather than switching abruptly. This addresses a known weakness of many existing unified control concepts, which rely on explicit mode switching (Lombaerts et al., 2020a; Angelov and Holzapfel, 2021; Dollinger et al., 2021). A continuous blending approach was introduced in earlier work (Angelov and Holzapfel, 2022), and the DLR team takes it further by giving the active side stick a role in guiding the pilot through the transition.

What They Found

The contribution of this paper is not a single headline number — there is no dramatic efficiency gain or breakthrough in hover performance. It is a demonstration that an integrated, unified control concept can be implemented cleanly and flown by pilots in a realistic full-motion environment, and that it imposes surprisingly little cost.

The paper's most concrete results come from two sources: piloted simulation on the DLR PAVSIM motion simulator, and an optimal-control-based analysis method borrowed from earlier trajectory-optimization work (May et al., 2025). This second tool is worth dwelling on because it is methodologically clever. The researchers wanted to know whether the command filters — which smooth and shape pilot inputs before they reach the flight control system — impose a meaningful performance penalty. The filters inherently limit how aggressively the aircraft can respond. To quantify this, they formulated an optimal control problem: find the fastest, most efficient way to execute a given mission, first with the command filters in place and then in a pure closed-loop system free of those filters, and compare the resulting mission durations and inceptor activity.

The finding is reassuring. According to the study, the command filter does not significantly increase mission duration compared to the closed-loop system — the cost of smoothness, in other words, is small. The researchers also found that the active side stick helps reduce inceptor activity: because the stick pushes back with forces that inform the pilot how much input is needed, pilots make fewer, more deliberate control movements rather than hunting and correcting.

The force feedback itself is the most intuitive and, in some ways, most compelling result. During specific flight phases — particularly the transition, where the aircraft is neither cleanly hovering nor cleanly cruising — the active stick generates tactile cues that guide the pilot. The stick doesn't take over; it nudges. It communicates, through the language of force, the character of the maneuver being flown. For a non-expert pilot, this is the difference between reading a flight instrument and feeling the aircraft's state in your hand.

Why This Changes Things

The significance of this work is best appreciated against the historical arc of vertical flight. The XV-3 and XV-15 taught engineers that tilt-wing aircraft were physically viable but demanded extraordinary pilot skill (Thomason, 1990). The Harrier taught the same lesson more forcefully and more publicly. Fly-by-wire systems then decoupled the pilot from the physics — the V-22 Osprey and the F-35B demonstrated that a unified control interface could make vertical flight routine for trained pilots (Robinson et al., 1989; Walker and Allen, 2002). But every one of those solutions assumed a professional aviator on the other end of the controls.

The DLR concept is aimed squarely at the civilian future. If eVTOL aircraft are to participate in the Low-Altitude Economy and Advanced Air Mobility — the umbrella terms for the coming ecosystem of urban air transport — they will be flown by a much broader population. EASA's draft certification rules explicitly contemplate this, and the Simplified Vehicle Operations framework exists precisely to lower the training and workload bar (Wing et al., 2020; Federal Aviation Administration, 2024; European Union Aviation Safety Agency, 2023).

This is where the active side stick quietly does the heaviest lifting. A passive stick — even a well-designed unified one — still requires the pilot to hold the aircraft's states in working memory. An active stick externalizes part of that load into the pilot's hand. The paper frames this as reducing cognitive load and improving situational awareness (Çağrı Ege Altunkaya et al., 2024), and the reduction in inceptor activity is evidence that the tactile channel is doing real work.

There is also a philosophical point embedded in the design, and it is refreshing. The researchers chose to keep attitude control available in hover and transition even though Simplified Vehicle Operations often renders it obsolete, because some pilots find direct attitude control useful. The main command is always the translational command — the pilot says "move this way at this speed," not "bank this angle" — but the system does not insist on a single rigid philosophy. It blends task-oriented and attitude-based control pragmatically, in service of the human rather than the concept.

What's Next

The honest limitations are stated plainly in the paper, and they are worth taking seriously. The flight dynamics model is reduced-order, stripped of sensor latency, wind, turbulence, failures, and other off-nominal conditions to isolate the core control concept. Real eVTOL flight will not be so kind. The pilots in the study are operating in a simulator with a reduced-order model of a real aircraft; the gap between that and a production airframe with the full mess of real-world disturbances is significant for any control law, and this one is no exception.

The optimal-control analysis, meanwhile, is a promising tool for clearing pilot control concepts, but it is a tool in early validation. The researchers propose it as a metric to rate and compare control concepts — a way to quantify performance penalties and inceptor activity before expensive piloted testing. That is a genuinely useful methodological contribution, but it will need broader application and cross-validation against more flight-test data before it becomes a standard practice.

Several questions remain open. How does the concept hold up when a real tilt-wing aircraft responds with the full fidelity of aerodynamic and structural effects the reduced model omits? How does pilot performance vary across a population with diverse training backgrounds — the paper validates the concept but does not yet report population-level workload or error statistics? And what happens when things go wrong: the design requirement that releasing the stick continues the current flight state is elegant, but off-nominal behavior, sensor degradation, and actuator failures will eventually need to be part of the picture.

What this paper offers, at its core, is a template. It demonstrates that a unified control concept with an active side stick can be built, flown, and evaluated in an integrated framework — and that the cost of making flying accessible is small, measurable, and worth paying. The three-hand problem belonged to the Harrier era. The argument here is that the next generation of vertical flight can be built on one hand, guided gently by a stick that knows the aircraft better than its pilot ever needs to.