Volume 01 | Issue 07 | July 20, 2020
Human-AI Trust, Cloud Designing & Cybersecurity

Today’s issue focuses on Connectivity (more specifically cybersecurity), Advanced Manufacturing and Autonomy & Artificial Intelligence.

Hope you enjoy. Thanks for reading.

— The Future of Aerospace Team
Machine Learning: Competency Awareness & User Communication
Greater incorporation of autonomy into aerospace will require more understanding of how autonomous systems function, both for certification and for effective partnership between these systems and their human users.

Machine learning, commonly used for pattern analysis and recognition, plays a major part in artificial intelligence for autonomy systems, particularly in pairing with sensors and cameras (i.e. obstacle avoidance, surveillance, search-and-rescue).

A DARPA program called Competency-Aware Machine Learning (CAML) is focused on improving our understanding of how machine learning systems — commonly used for sensor data pattern analysis and recognition, such as for obstacle avoidance and search-and-rescue — work, along with recognizing when they are operating in unfamiliar contexts and communicating that to users.
  • CAML’s objective is less about certification and more about providing operators insight into the capabilities of a given machine learning system, for example to assist in choosing the right ML assets for a task, and offering real-time alerts when a task moves outside of a system’s competencies and user intervention is needed.

  • Since ML systems acquire their task-specific capabilities by analyzing lots of data, the CAML program will augment the ML system during the learning stage by enabling it to both learn how to do the task and understand the conditions in the data under which it performs well or poorly. This is accomplished by keeping track of scenarios the ML system encounters and finding a way to tie the scenario with how well it performs the task.

  • BAE Systems, one DARPA-contracted company for CAML, recently delivered the first iteration of its software solution, called MindfuL. During the planning stage of a mission, the system communicates competency scores for various ML systems relative to tasks, and then during real-time performance describes environments it encounters as “familiar” or “unfamiliar” to the underlying ML system and recommends user actions.
DARPA’s four-year program aims to develop general competency-aware capabilities that can be incorporated into all ML systems, with potential applications including autonomous navigation and obstacle avoidance, UAV search and rescue, autonomous ground resupply vehicles, self-driving taxis and missing planning. Targeted platforms include unmanned air and ground systems as well as robots.

BAE’s current focus is on image classification and object identification, and the company is pursuing application to ground vehicle navigation, leader-follower convoying, route planning, and target recognition through satellite imagery.

One natural application of this type of system, according to the company, would be identification of degrading sensors, as environments will appear as “unfamiliar” when input data is compromised and therefore bears less resemblance to the data an ML system was trained on.

Read more on how to understand the competencies of ML systems and build human-AI trust.
Designing Aircraft in the Cloud: A Growing Trend for Aerospace Engineers
Every aspect of next generation aircraft design can be accomplished today in the cloud, with engineers across hundreds of aerospace industry suppliers now using a common new set of all-digital tools in their development efforts to help advance manufacturing efficiencies.

Engineers have been using computer aided design tools for decades to develop complex structures, components and systems.

Now, the cloud has transformed the way these tools are managed on an interchangeable basis between parts suppliers, systems integrators, software developers and other stakeholders within the design and manufacturing ecosystem required to actually build a new aircraft, engine, system or component.
  • Dassault Systèmes, a Dassault Group subsidiary, claims that every major aircraft maker today, ranging from Airbus, Boeing and Gulfstream to Lockheed Martin is using their digital design tools for the development of new aircraft.

  • Research on aircraft development workflows collected through observations of some of these customers in recent years by the company found that in the new aircraft development process, engineers spend about 30 percent of their time searching for the data that they need to work on through multiple databases, with no digital continuity.

  • Traditional searches observed were mostly restricted to “full text” or “attribute” based search with limited capability to filter against images or videos. In 3DExperience, the search option is much more powerful as it knows a lot about the semantics of engineering, including lifecycle, configuration and three-dimensional spatial zones.
Frédéric Chauvin, A&D Industry Solution Experience Director, Dassault Systèmes told Avionics: “From requirement to validation and verification. From bolt to aircraft. From concept operations. Cloud is not restricted to small or niche scenarios.”

Some of the aerospace industry’s largest and emerging names are using 3DExperience in the cloud to develop next generation aircraft designs.
  • Joby Aviation, one of Uber Elevate’s air taxi development vehicle partners, as a company that is using 3DExperience to address weight reduction and aerodynamic simulation associated with their future facing air taxi design.

  • Boom Supersonic, whose XB-1 subscale supersonic demonstrator is to rollout of its hangar in October, used 3DExperience for structural analysis and systems architecture development.

  • In June, Vertical Aerospace, a startup air taxi maker based in the U.K., announced plans to adopt the 3DExperience platform on the cloud to develop their all electric air taxi.
Chauvin: “An example on the simulation side is complex aerodynamics studies as well as structure, and other system simulations. It requires powerful processing capabilities that are now hosted on the cloud. You can launch the [computational fluid dynamics] CFD analysis from your mobile or tablet and get notified when the result is available. Latency and bandwidth are carefully handled to provide best performances for any situation. We have customers developing complete aircraft on the cloud with 4G cellular networks. They don’t even feel the need for high speed fiber network.”

Read more on how cloud-based design tools are changing aerospace.
Aerospace Cybersecurity Moving Away from Network-Based Security
Artificial intelligence and data defense at the application level, rather than the network level, may feature prominently in the future of aerospace cyber security.

In the United States, a so-called Zero Trust Architecture (ZTA) has been gaining traction, which assumes networks are compromised and instead focuses on the defense of applications’ data.

Air Force Brig. Gen. Chad Raduedge, director of cybersecurity and information dominance for Air Combat Command (ACC), told us:
  • “The Air Force is aggressively pursuing a Zero Trust strategy and our senior leaders have recognized the necessity to change how the DoD operates in and through cyberspace to counter an increasingly effective cyber adversary.”

  • “Zero Trust concepts enhance the Air Force’s ability to protect our data, while also enabling mission agility and mobility. Each of the DoD Service Cyber Components, including Air Force Cyber (AFCYBER), are developing and executing Zero Trust pilots to refine strategies intent to rapidly implement Zero Trust architectures and strategies across the Air Force.”
Air Force Lt. Gen. Timothy Haugh, in a virtual discussion last week: “From a defensive perspective, we want to rapidly transition from network edge defense to data defense … That changes a number of things: our architecture, how we approach it, how we train airmen, and also which things we can automate and which things we need to still defend with the human to be able to translate that.”

The House version of the annual U.S. defense spending bill, which has yet to be finalized, includes language encouraging the Secretary of Defense to “implement a Zero Trust Architecture to increase its cybersecurity posture and enhance the department’s ability to protect its systems and data.”

On AI and cybersecurity: NATO has been looking into an Autonomous Intelligence Cyberdefense Agent (AICA) approach, believing that centralized, human-directed cyber defense will not be feasible in future cyber conflicts.

Paul Theron, a cyber resilience scientist with Thales and member of NATO’s research group on this subject, believes that the future of cyberattacks “is likely to rely on autonomous intelligent cyber-weapons,” and therefore an autonomous cyber defense is required.

“Goodware will fight malware.”

Read more on the future of aerospace cybersecurity.
Thank you for reading the Future of Aerospace, brought to you by Avionics International.

View in web browser

This message was sent to lhofmann@accessintel.com

Access Intelligence LLC * 9211 Corporate Blvd., 4th Floor, Rockville, MD 20850
Unsubscribe | Privacy Policy | Contact Us