This paper presents a global intelligence system that synthesizes aerial vehicles’ real-time physical data, planned actions, and historical behavior into engineered data frames representing the collective state of the airspace and suitable for efficient machine learning consumption. These data frames are then learnt by a deep neural net to build a prediction model that estimates the expected evolution path of the current state, thereby identifying potential future conflicts. This approach lends itself to an automated early warning system that the aerial vehicles can implement onboard with a suitable edge computing module more efficiently and effectively than non-AI methods, and eventually take preventive or corrective measures towards self/collaborative resolution of the issues. Contrary to a centralized early warning system where all vehicles’ task-space eventually converges to a global optimum state, the presented distributed global intelligence system brings in a balance between local utility functions of each vehicle and the global operating framework. This contributes to effectively handle the potential massive scaling in urban air mobility in the near future.
AEGIS: Autonomous Entity Global Intelligence System for Urban Air Mobility
Aviation 2020 ; 2020 ; Virtual, US
Aufsatz (Konferenz)
Keine Angabe
Englisch
Getting AEGIS to Sea: The AEGIS Ships
Online Contents | 2009
|Getting AEGIS to Sea: The AEGIS Ships
Online Contents | 2009
|Online Contents | 2009
|