- January 5, 2016
- Posted by: Jeff Kish
- Categories: AI & Advanced Computing News, AI Transportation Systems News, Complex System Analysis News, Latest News, Modeling, Simulation & Visualization News, Research & Development News
The National Airspace System (NAS) generates a large amount of heterogeneous data from a variety of software and hardware systems. The SMART-NAS Test Bed will closely mimic the NAS operation and generate large quantities of multi-structured data including text, numeric data, and multimedia, from individual subsystems and participating individuals. With such huge, complex, and quickly growing data sets, it is daunting to quickly identify and explain the causes of anomalies and the failures that lead to undesirable end states or incidents. IAI and collaborators at Metron Aviation will leverage trusted Big Data technologies and scalable and tailored data mining algorithms, that can be applied to a variety of large NAS data sets with other performance data. Solutions will be found by identifying anomalous patterns hidden deep within the data sets. The team will extend the capabilities of IAI’s in-house Big Data storage and analytics tool, ATLAS, for Air Traffic Management (ATM) researchers. ATLAS is a multi-level, multi-objective tool that extracts and manages ATM data from a variety of ATM data stores and prepares it for holistic analysis within an actionable time line. The ATLAS architecture allows for deployment on private or public cloud infrastructure, and can be accessed via a website that acts as the ATLAS portal for the end user. Scalable data mining and machine-learning algorithms, especially tailored for large complex SMART NAS data sets, will be included in ATLAS to help in the diagnostics and prognosis of data patterns of interest to NAS stakeholders. These algorithms will be integrated with SMART NAS data sets and available ATM big data archives, enabling the tool to learn to mimic user interactions with data, and to isolate interesting use cases more effectively than with current rule-based algorithms.