Microscopic analysis of many optimizing air vehicles using high-performance computing
Author(s)
Sanni, Olatunde B.
Advisor(s)
Feron, Eric M.
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The operational success of an air transportation system (ATS) depends on air traffic policies. These policies balance the trade-off between safety and performance. Stringent policies stifle rewards, and lenient policies can lead to loss of life and property. Air traffic management (ATM) research explores this trade-off. Unsurprisingly, this research area has been limited by the human-in-the-loop because human pilots and air traffic controllers are difficult to predict and expensive to model. However, advancements in autonomy algorithms and computational systems are changing this landscape. This dissertation leverages these advancements to explore consequences of air traffic policies. A novel approach for studying air traffic policies is presented. This novel approach is implemented in the developed Massive Air Traffic Simulator (MATS). This approach uses a high-performance computing (HPC) cluster to conduct real-time simulations of many autonomous and independent air vehicles that communicate with ATM services, such as a weather service, building service, and traffic control or deconfliction service. Its simulated air vehicles use trajectory optimization to account for air traffic policies. The optimizing air vehicles in this dissertation use the developed Extensible Trajectory Optimization Library (ETOL), which transcribes an optimal control problem into a problem that path-planning and optimization software can solve. These vehicles asynchronously use ETOL as part of a model predictive control (MPC) strategy, and ETOL is configured to use a mixed-integer linear programming (MILP) solver. Although trajectory optimization is notorious for taking a significant amount of time, this dissertation demonstrates the developed approach’s real-time worthiness. This dissertation provides three primary contributions. It presents a novel framework for assessing large-scale air traffic operations at real-time speed. It presents a MILP formulation that air vehicles can rapidly solve as part of a MPC strategy. It presents simulation results, lessons learned, and challenges for large-scale high-density air traffic.
Sponsor
Date
2023-01-13
Extent
Resource Type
Text
Resource Subtype
Dissertation