I spend many hours in a flying laboratory where I take a Sikorsky S-76 helicopter through flight tests, though my hands rarely manipulate the flight controls. Instead, I plug the flight destination and mission requirements into a tablet computer resting on my knee in the cockpit, and the aircraft does the rest. My hands are by my side, and my trust is in the machine.
After 22 years as a U.S. Coast Guard search-and-rescue pilot and 10 years as a Sikorsky test pilot, today I’m the lead development/project pilot for Sikorsky Innovations Autonomy programs. I’ve logged more than 300 hours of autonomous flight in the Sikorsky Autonomy Research Aircraft (SARA), an S-76B helicopter we’ve modified with our Matrix Technology that enables two, one or zero-pilot flight.
You read that right. I’m a retired Coast Guard pilot who spent years executing complex missions including saving lives, and I now fly an optionally piloted helicopter. I believe this technology will enable even more critical lifesaving missions, just as Sikorsky founder Igor Sikorsky dreamed the helicopter would do. To understand why I’m so passionate about how Matrix will evolve, let me take you back to one particularly dark and stormy night.
A classic nor’easter was in full swing, with one-quarter-mile visibility in blowing snow, 200-ft ceiling, a brisk wind of 45 to 60 kt and seas of 20 to 30 feet. The distress call came in from a 47-ft fishing vessel with five fishermen on board at 3 a.m., Jan. 11, 2000. The boat was 30 miles off the North Carolina coast and taking on water. That night, my Coast Guard crew performed at or above peak performance for nearly five hours. At the end of the mission, we had nothing left in the tank, both literally and figuratively. If we went into that rescue a little more tired, distracted or a little less prepared, the results could have been very different. Those five fishermen may not have made it home to their families.
I think about that rescue when I’m asked why I would trust a machine to complete demanding missions with a reduced crew or even with no crew at all. I don’t hesitate to explain: We allow the machine to do what it does best and allow the human to do what it does best. Sikorsky calls it “pilot-directed autonomy.”
Flying with Matrix
The concept of autonomy runs the spectrum from robotic assembly lines to robotic vacuums and is generally aimed at replacing the human. Sikorsky’s definition is different. We put the human center stage. Instead of replacing the pilot, we are taking an evolutionary step to augment a pilot’s ability to operate in an environment that has increasing demands and complexity. Pilot-directed autonomy provides operators with a transparent and pervasive layer of technology that will reduce their workload in all phases of flight.
Matrix is a suite of technologies that includes external sensors, high performance computers, actuators, advanced proprietary flight control laws, path planning algorithms and pilot interfaces. It is designed to be a drop-in replacement on existing aircraft.
Early in the development of the Matrix technology suite, Sikorsky saw close parallels with an emerging Defense Advanced Research Project Agency (DARPA) program, the Aircrew Labor In-cockpit Automation System (ALIAS). With Matrix as its baseline, the ALIAS program’s goal is to introduce high levels of automation to enable operation with reduced onboard aircrew while improving mission performance and safety. DARPA awarded Sikorsky its first contract for the ALIAS program in 2015 and we are now working in partnership executing the third and final phase of ALIAS development.
My “office,” SARA, is a standard 11,300-lb Sikorsky S-76B helicopter outfitted with the Matrix and ALIAS technology suite. S-76 pilots and passengers know the aircraft from providing 30 years as a safe and reliable medium utility aircraft used worldwide.
SARA is not a remote-controlled aircraft nor an oversized drone. It can leverage full-authority flight control inputs for autonomous flight — including takeoff, route planning, obstacle avoidance, site selection and landing. Sensors give SARA a 3D understanding of its environment in real time.
In designing our SARA interfaces, we attempt to provide the pilot with the maximum amount of bi-directional informational flow with the system and the minimum amount of input. Think of it as flying the machine at the mission level versus the task level. Rather than inputting individual waypoints as with a traditional flight management system, you instruct the system to simply “take me to position X and do Y when we get there.” All of the resultant flight planning is editable either before the flight or in real time while flying.
Our baseline hardware for this interface was a tablet, mainly because it was portable and easy to program. I can recall the first time I commanded “takeoff” with the tablet. To my surprise (and a small bit of reluctance), it did exactly that. Tablet control represents the complete autonomy end of the control spectrum. We’ve since been experimenting with other control hardware that allows the pilot to control the machine throughout the autonomy spectrum.
I have flown to outlying airports in SARA, never once having touched the controls. SARA has taken me hands-free into a hover over a moving simulated ship deck in the Long Island Sound. I have logged well over 100 autonomous takeoffs and landings and over 300 hours of autonomous flight. These are extreme examples of reducing pilot’s workload to essentially zero. Sikorsky has been flying SARA for five years and the experience has helped shape our vision of pilot-directed autonomy.
The Potential of Pilot-Directed Autonomy
Just as a flight director overlays an autopilot system, which in turn overlays a stability augmentation system, we view pilot-directed autonomy as an overarching layer in the control-system path.
It will always be present and will never be intrusive. The level of automation will always be up to you, the pilot. You can choose along a spectrum from full mission execution to simply having it run in the background. This choice would be dictated by the complexity of the mission and be directly proportional to the level of workload you desire.
The focus of DARPA’s ALIAS program is to allow operation of multi-pilot type military/commercial-certified/qualified aircraft with reduced crew. For this to become a reality, we are testing in both fixed- and rotary-wing aircraft. We anticipate the need to demonstrate an autonomous land-safe capability in the event of pilot incapacitation. This scenario is in line with the full spectrum capabilities of the Matrix system.
Sikorsky’s experience on full authority fly-by-wire systems on the CH-148 Cyclone, the CH-53K King Stallion and the S-97 Raider has shown us how to integrate advanced external sensors into the control path and demonstrate baseline capabilities like obstacle avoidance and landing zone site selection. We have discovered that the real technical challenges lie in the seamless integration of these capabilities with the human operator. Human-machine interface design is a large focus of our efforts and a key component to obtaining the appropriate levels of trust that will be required for our efforts to be successful.
The Matrix system is best thought of in terms as a “digital co-pilot.” The algorithms we have designed (and are designing) are tailored to the aircraft and are intended to encapsulate all systems, functions, performance and limitations. Some heuristics can be built in, such as diagnosing a system malfunction and proposing the appropriate emergency procedure, but the pilot in command will always have the final say before critical systems are shut down. This is intentionally modeled after the pilot in command/second in command relationship you see in human-crewed aircraft. Most of us know this as “dual concurrence.”
As a result of our partnership with DARPA, we soon will have another aircraft flying Matrix technology: a full fly-by-wire UH-60 Black Hawk. We also have plans in place to augment the S-92 aircraft’s flight control path with Matrix. Our Matrix developments have been driven by the need to obtain civil certification and military qualification. DARPA, our military partners and the FAA have been involved with our progress since day one.
When integrated properly, pilot-directed autonomy can mitigate critical pilot-related causes of accidents like controlled flight into terrain or drastic unusual attitudes associated with inadvertent instrument meteorological conditions. It can have a default failsafe capability that will monitor pilot inputs and compare them against mission goals and real-time environmental constraints. It could “get in the loop” and either notify the pilot or, in extreme cases, disallow input, depending on the level of autonomy that the pilot selected. It could have background subroutines solely dedicated to monitoring its own health and be able to take appropriate and immediate action in the event of a system failure. For example, it could be constantly calculating a best-case autorotation landing site. It would do this at multiple times per second and always be ready to apply that all-important collective input and first turn. It could even complete the autorotation to touchdown if you so choose or maybe physically can’t. We have tested this in high-fidelity simulation, and it works.
Imagine being able to navigate in difficult weather over terrain, take a brown-out-prone landing zone, land on an oil rig in zero-zero weather, or hover over a fishing vessel in gale force conditions and never register more than a resting heart rate. That’s the future Sikorsky envisions with pilot-directed autonomy for our future and legacy aircraft.
We are not replacing the pilot; we are enabling him or her to do more and do it with an even larger margin of safety. RWI