Pilots are expensive to train and employ, and the long-forecast pilot shortage is actually materialising in many parts of the world including the USA. Logic suggests, therefore, that if robotic systems could replace one of the pilots in a two-crew aircraft, or even carry out just some of the co-pilot functions so as to make single-pilot operation feasible, they would definitely have a commercial future.
Viginia, USA-based Aurora Flight Sciences is working with the US Defense Advanced Research Projects Agency (DARPA) to develop robotics that can be applied to aircraft operation. DARPA itself is a world leader in robotics for the military and emergency services, but Aurora’s aerospace expertise in drone development and manufacture make it an ideal partner for extending robotics into aviation operations, including commercial air transport and general aviation as well as the military, and rotary-wing as well as fixed-wing.
The company has already done demonstration flights with their robotic system in charge of a Boeing 737-800NG simulator and an airborne Cessna 208 Caravan, of which videos are available on the Aurora website. The 737 trial was carried out at the John A. Volpe National Transportation Systems Center in Cambridge, Massachusetts, demonstrating ALIAS’s capability to manage the existing 737 autopilot and auto-landing system to guide the aircraft to a safe landing, which Aurora says shows what it can do in the event of pilot incapacitation.
The company points out it has also tested ALIAS components successfully on an airborne Diamond Aircraft DA42 light twin, a Bell UH-1 Iroquois helicopter, and Bombardier DHC-2 Beaver, which – with the 737 – pretty much covers the entire gamut of cockpit types from digital to clockwork. The DA42 trial was the most recent, and in it ALIAS was used, according to Aurora, without intervention by the safety pilot, to demonstrate management of the aircraft’s approach and landing procedure to a fully automated landing, except this was simulated on a “virtual runway” set at 3,000ft.
If the idea of a humanoid robot strapped into the cockpit right hand seat sounds as if it might pose a major psychological challenge for today’s aircraft commanders, they can take at least a little comfort from the fact that the visible mechanical devices have nothing humanoid about them. They consist of smart, multi-jointed manipulator arms, and their motor unit is not strapped into the co-pilot’s sheepskin-covered seat – the latter is entirely removed. The flight controls are manipulated by a rod system directly connected to the yoke and rudders, and the manipulator arm, mounted close to the centre console, operates switches, knobs, pushbuttons, and levers like those for the throttles, flaps, gear or spoilers. The manipulator arm, says Aurora, is a commercial-off-the-shelf system with a multi-functional “hand” that can perform precise actions.
The question is, since autopilot systems already involve invisible robotics and they have worked well for many decades, why develop a robot to control the aircraft from the cockpit when the autopilot can already perform that function?
Depending on the ALIAS capability the customer chooses, the system can fly the aircraft by directly manipulating the primary flight controls as a pilot does, or it can physically manipulate the knobs, pushbuttons and switches on the autopilot flight control panel (FCP) – also just as a pilot does – to direct the autopilot.
The advantage to the aircraft operator is that the robotics that Aurora is developing in its ALIAS programme (Aircrew Labour In-cockpit Automation System) can be applied to existing aeroplanes without the need for modification to the aircraft’s avionics, software or control systems. Thus the operator will – potentially – be able to reduce crew numbers by 50% quite quickly.
Aurora Flight Sciences
As ALIAS programme manager Jessica Duda points out, in the same way that a pilot needs to take some days or weeks to gain a type rating on a new aircraft, an ALIAS robotic system can be re-programmed and modified in about 30 days to fly any aeroplane. The company explains: “Custom-developed software includes a reconfigurable set of plug-ins for the wide variety of instruments and effectors that are found in cockpits: switches, lights, knobs, levers, gauges, etc; and these plug-ins can be rapidly tailored to the specific layout, number, and type of instruments and effectors in a new cockpit.” Also built in to the system is “a knowledge-acquisition process that facilitates transition of the automation system to another aircraft within a 30-day period.”
Aurora explains this rather more bluntly in the ALIAS brochure: ALIAS, it says, is “a tailorable, drop-in, removable kit that would promote the addition of high levels of automation into existing aircraft, enabling operation with reduced onboard crew.”
Aurora emphasises that ALIAS is designed as an assistant for the aircraft commander, not the thin end of the automation wedge directly leading to a completely pilotless commercial aircraft. ALIAS does not have any mission planning or management capability, nor sophisticated cognition as a human pilot does. Like an autopilot, it needs to be told the details of the task it is required to perform, then told to engage before it can begin the task.
ALIAS may not have sophisticated cognition like a human does, but there are still many parallels between its function and some tasks that a co-pilot would normally perform in a two-pilot cockpit. The main parallel is that ALIAS, like the pilots, gets its flight and systems data input “visually” from the flight instruments, and from the engine and systems instruments. Its visual input comes from “machine-vision” cameras that scan the instrument panel, like a very advanced version of number-plate recognition. In Aurora’s words: “Machine-vision cameras, trained on the instrument panel, non-invasively perceive the aircraft state.” Since it can be programmed to recognise regular flight procedures associated with particular phases of flight, it can perform a useful monitoring function, like recognising that the gear has not been selected down when it should have been.
The commander-to-robot interface is not socially intuitive, like a tap on the shoulder, an exclamation or a pointed finger might be for a human co-pilot. Interface with ALIAS is via an in-cockpit tablet used by the aircraft commander to process and monitor checklist actions and other functions. For example the tablet also accepts speech commands, feeds back synthetic voice responses and enables the pilot to manage the distribution of tasks between the human and the robotic systems. ALIAS can, of course, be disengaged at any time, like an autopilot.
After about two years in development, ALIAS is still in its prototype stage despite some successful early trials, and Aurora is reluctant to forecast how long it will take before it is anywhere near a production version. Indeed it has not yet proven the ability to cope with all phases of flight from take-off to landing. John Wissler, Aurora’s vice-president of research and development explains: “As we move towards fully automated flight from take-off to landing, we can reliably say that we have developed an automation system that enables significant reduction of crew workload.” To balance Wissler’s remark, programme manager Jessica Duda comments that ALIAS was not conceived as a take-off-to-landing piloting system, which might be taken to imply that co-pilots are not necessary all of the time. Some people might not argue with that. But Duda’s message is that ALIAS should be capable of being a pilot’s assistant throughout all phases of flight.
Co-pilots, however, are traditionally accepted as having a monitoring role as well as a “doing” function, although Aurora points out the oft-quoted mantra that humans are notoriously bad at monitoring for any length of time. Acknowledging this, the company says it is also “working on a version of the system without robotic actuation that, instead, aims to support the pilot by tracking aircraft physical, procedural, and mission states, increasing safety by actively updating pilot situational awareness.” The question here is, if there is to be no robotic actuation to enable physical intervention by ALIAS, that cuts out one of the layers of safety that a real co-pilot can supply. According to studies by Toulouse-based French aerospace research agency ISAE-Supaero, a pilot flying who is already overloaded can block out warnings, whether delivered by audio alert, voice, or written advice on the warnings display. So the alert is going to have to be intuitive enough to cut through overload-generated pilot freeze.
Aurora says the pilot can choose ALIAS’s role according to his or her needs: “ALIAS can operate as a flight monitor by monitoring the state of the aircraft and comparing it with known flight phases, as well as retrieve procedures and record flight data.