A key component of sixth generation aircraft system of systems (NGAD, FCAS, Tempest/GCAP, etc.) will be the deployment of Collaborative Combat Aircraft (CCA). There are however outstanding challenges that are often lost in the hype:
-Airworthiness certification standards for truly autonomous aircraft with learning-enabled systems (LES) such as those built with deep learning and reinforcement learning algorithms.
-The human factors (including the neuroergonomics) of effective Man-Unmanned Teaming (MUM-T).
-The development of new Tactics, Techniques, and Procedures (TTPs) for operationalizing CCAs.
Expecting pilots of crewed aircraft to control uncrewed CCAs could significantly increase their already high workload and reduce their cognitive performance, so these CCAs will need to have a high level of autonomy with goal-directed control (as opposed to today’s remotely piloted aircraft systems or RPAS). Humans will be setting those goals according to mission objectives but in compliance with international laws, norms, and regulations. So, there is a need for what developmental psychologist Michael Tomasselo called "shared intentionality" between human pilots and the CCAs. There are also operational benefits in a distributed approach granting more autonomy to CCAs.
Learning Enabled Systems (LES) such as those built using deep learning and reinforcement learning introduce new testing, verification, and certification challenges resulting from difficulties in providing formal guarantees of system behavior. In the aviation domain, certification standards for LES are still a work-in-progress and a subject of ongoing academic research.
A CCA costing millions of dollars will certainly be required to comply with a minimum level of certification standards. For example, the European Military Airworthiness Certification Criteria (EMACC) Guidebook specifies standards and processes for military airworthiness certification in Europe. In civil aviation, the ARP6893/ED-324 Machine Learning standard by the joint EUROCAE WG-114/SAE G-34 working group is still a work in progress.
Aerospace engineering practices and methods must evolve to support learning-enabled, autonomous, adaptive, high-dimensional, highly non-linear, partially observable, and non-deterministic system behavior under uncertainties. Explainability is a sine qua non in the aviation domain.
New hazard analysis methods that account for interaction between system components including software, hardware, and human factors are needed (Systems-Theoretic Process Analysis or STPA is an example). Safety challenges can be addressed through Formal Methods, Uncertainty Quantification, Safety Filtering, Run Time Assurance (RTA), Reachability Analysis, Fault Detection, Isolation, and Recovery (FDIR), Dynamic Fault and Attack Injection, and Scenario-Based Testing.
#ai #aviation #aerospace #aircraft #ucav #airforce #defense #war #warfare
Leading the Delivery of Next-Generation Integrated Training Solutions
3wWelcome to the Project OdySSEy team 💪 a fantastic addition and we have already really enjoyed how we have collaborated together, long may it continue!