Mercury VP Sees EVTOLs Driving Future of Avionics Embedded Convergence
As electric air taxi programs continue to proliferate and advance their designs, the ambitions of speed of development from companies behind them will drive the evolution of embedded safety critical avionics computing and processing architectures.
The use of an electrically-powered air taxi to commute between vertiports in traffic-choked cities might seem like a futuristic fantasy, but one of the world’s largest suppliers of open systems-based embedded avionics is already working on safety-critical computing architectures with several companies that are developing electric vertical take-off and landing (eVTOL) urban mobility vehicles.
Ike Song, Mercury Systems vice president and general manager of mission systems, is excited about the future engineering prospects of the electric air taxi market and the development it promises to drive.
“There’s a huge growth in the number of legacy aircraft that needs to be updated from a DoD perspective both domestic and international, but what’s even more exciting from a future engineering perspective is the eVTOL arena,” Song sand. “There are hundreds of development programs right now and they’re talking about fielding thousands of these things rather quickly.”
A December 2018 white paper published by Song entitled Digital Convergence outlines what Mercury’s engineers believe could enable the future safety-critical, secure architecture for eVTOLs and other aircraft. The paper proposes a logical mapping of an aircraft’s sensor chain to its mission processing grid. According to the white paper, the average in-service air transport aircraft can have up to 15 discrete sensor- and mission-processing subassemblies on it.
Mercury wants to place that federated architecture of digitally converged sensor- and mission-processing functionality into more integrated boxes in the future. In effect, the concept of digital convergence creates sensor-processing subassemblies with real-time processing engines that turn data streams from sensors and electro-optical focal planes into actionable information.
One of the unique approaches to embedded processing architecture that Mercury has pursued to optimize size, weight and power is the use of three-dimensional stacking, according to Song.
“We call it wafer stacking,” Song said. “As an example, for the more predictable-sized packaging of components such as memory, we purchase that from companies such as Micron and instead of integrating them on boards next to each other as we might have in the past, we’re stacking them on top of each other using a unique and robust approach to precise wire bonding fabrication to enable a decoupling of the devices that are being supported.”
Mercury is already combining the use of wafer stacking and system-in-package miniaturization technologies with an expansion in mezzanine site sizes in new processing configurations. The latest example of this was used to create its new EnsembleSeries LDS3517 processing blade. LDS3517 combines Intel Xeon D server class processors and Xilinx’s field-programmable gate array (FPGA) components into a 3U OpenVPX form factor. It further combines the computing power of three circuit boards into one with a connector capable of supporting a wide variety of applications, including connection to a chassis or a graphics processing unit (GPU).
The company is also forming key partnerships with other providers of embedded architecture designed for eVTOL companies. Mercury’s partnership with Zürich-based startup Daedalean is one of the most promising. Daedalean describes its mission as developing autopilot software capable of passing the exams of human pilots. Its software is capable of drone and wire detection, three-dimensional mapping and autonomous landing. At the 2018 Farnborough International Air Show, the two companies did a live demonstration running Daedalean’s software on Mercury’s Rock-2 modular subsystem chassis.
That is the type of consolidated processing architecture that could enable the type of artificial intelligence and machine learning that will be required for navigating eVTOLs in the future.
Can Rapid eVTOL Evolution Drive Digital Convergence?
In order to enable the creation of new avionics functionality on next-generation commercial and military aircraft, companies such as Mercury Systems have to invest in the acquisition of new computing chips, storage components, memory devices, backplanes and other embedded circuitry, wiring bonding and processing components.
For the moment, the traditional refresh cycle for avionics systems on military aircraft – every two to four years – will continue to drive much of that development, while commercial and business jet avionics refresh cycles increasingly rely on more flexible software-defined architectures inside of systems refresh timelines spanning 10-15 years or more.
The volume of demand for new functionality is relatively low, especially when compared to the total quantity of industrial internet-of-things and autonomous vehicles that also need safety-critical embedded computer processing capabilities.
But the eVTOL market is rapidly expanding to the point where it could take the lead in demand for new functionality requirements. Right now, according to the latest numbers published by the Vertical Flight Society, the promise of urban air mobility is driving the development of nearly 150 eVTOL aircraft design concepts.
“When you look at the U.S. military for example, their biggest in-service fleet is really a combination now of different Black Hawk variants and the F-35 as that starts to enter into service,” Song said. “When you look at the eVTOL market, they’re anticipating hundreds of thousands of these aircraft within a very short period of time. They’re moving at a much faster rate than the average avionics refresh cycle for traditional commercial and military programs. Not only a faster rate from the kind of processors that they need, but also technology that they want to embed.”
A 2018 study published by Porsche on the type of ground infrastructure and aircraft design needed to enable the future use of electric air taxis places the aircraft design concepts of future electric air taxis into three categories. The first includes multirotor systems, such as the air taxi being developed by German startup Volocopter, which distributes multiple motors around the periphery of the airframe. The second category of design concepts uses both a lift and a cruise configuration, such as the Aurora eVTOL, which combine rotors for lift and fixed wings for forward flight.
The final category of market entrants, of which Lilium’s eVTOL concept is an example, relies on “tilt-x” designs in which wings, rotors, and ducts can be tilted to achieve the transition between vertical and horizontal flight. Flight control is accomplished by varying the speed of the individual rotors.
Boeing started off 2019 by completing the first flight of its own eVTOL prototype, using eight rotors to achieve vertical lift and a tail rotor for forward flight – all powered by an electric propulsion system borrowed from the X-24A’s sub-scale generator design.
Most of these designs still have not released many details on what their flight control systems, cockpit arrangements or safety-critical embedded computing architectures will look like. But there are companies that are working with Mercury and others, such as NVIDIA, to enable the type of processing that these aircraft will require in the future.
An air taxi in development by Airbus could provide more immediate insights on how the digital convergence of embedded avionics processing can support future functionality. Airbus has been flight testing its Vahana air taxi prototype at its Silicon Valley-based A3 facility since 2017.
Vahana’s onboard sensors and cameras will have to be able to detect objects at long distances and navigate around them autonomously. The French manufacturer’s California division believes GPU technology could help support the type of artificial intelligence and machine learning applications that the company is envisioning for a future certified electric air taxi.
The challenge Airbus faces in deploying the type of embedded processing needed to enable obstacle avoidance detection on the Vahana points to one of the first applications where the type of digital convergence proposed by Song in the Mercury Systems white paper can solve: enabling the use of a large number of cameras by a centralized flight control computer to navigate above and around vertiports and other flight environment obstacles.
“There are some eVTOL programs considering the use of up to 32 different cameras on their aircraft,” Song said. “That’s 32 different sensors that need a massive amount of processing power. Compare that to [self-driving car software developed by] Tesla and how many teraflops per second processing power are required for just six cameras on one of their models.”
NVIDIA’s Drive PX processors embedded into the current autopilot system on Tesla models features a multi-chip configuration with four artificial intelligence processors capable of producing up to 320 trillion deep learning operations per second.
Going beyond that type of processing power to not only support and enable sensors on an eVTOL, but also other safety-critical communications, navigation and surveillance functionality, will require a major step forward in the level of embedded processing available to avionics manufacturers today.
Right now, as the majority of eVTOL programs are still in their earliest stages of flight testing and development, the possibility of digital convergence enabling their safety-critical embedded architectures remains to be seen.
“The use of artificial intelligence, machine learning applications, smarter sensors — all of those are going to come together in the eVTOL arena,” said Song. They’re moving at a much faster rate than any other market segment right now and they don’t have a legacy architecture to worry about.”