Olatunde Sanni
(Advisor: Prof. Eric Feron]

will propose a doctoral thesis entitled,

Microscopic Analysis of many Optimizing Air Vehicles

using High-Performance Computing

On

Tuesday, July 6 at 10:00 a.m. (EDT)
BlueJeans: https://bluejeans.com/252729473/5871

Abstract
Path planning is essential in any guidance, navigation, and control (GNC) architecture. The human-in-the-loop typically performs this planning function; however, this approach becomes unsustainable as operational requirements increase. Consequently, algorithms are increasingly used to solve planning problems. Although many planning algorithms exist, their performance can significantly degrade in high-density air traffic management (ATM) problems that include many constraints, and this performance is the focal point of this thesis.

High-density air traffic is an active area of ATM research because it impacts many nascent aerospace operations, such as next-generation air transportation system (NextGen) operations and urban air mobility (UAM) operations. ATM research primarily examines safety and performance, and this work contributes a novel approach for gathering this data. This approach entails large-scale simulations of air vehicles that continuously optimize their own trajectories. This optimization is achieved with the Extensible Trajectory Optimization Library (ETOL), which we created to solve path planning problems with various trajectory optimizers. With ETOL’s interface to trajectory optimizers, simulated vehicles continuously generate optimized paths that account for local information from sensors and global information from traffic management service suppliers. Consequently, ETOL enables a possible elucidation of emergent behaviors that are byproducts of unforeseen flight conditions. This research also achieves a high level of fidelity with a high-performance computing (HPC) cluster that simulates thousands of optimizing air vehicles. This feat is achieved by using container technology to run a novel multi-process C++ simulation. This simulation uses Georgia Tech Research Institute’s multi-robot simulator, with the Robot Operating System (ROS) and ETOL, to visualize and analyze air traffic operations. At an ATM level, this research contributes a high-fidelity analysis of high-density air traffic operations under various flight rules. At a GNC level, this research contributes a review of ETOL along with its mathematical approach to decouple path planning problems from trajectory optimizers.

In this thesis, a series of experiments test the hypothesis that continuous trajectory optimization, within high-fidelity large-scale simulations, yields good insights about the safety and performance of high-density air traffic operations. In these experiments, thousands of optimizing air vehicles will autonomously perform tasks. As these vehicles execute their tasks, their performance will be captured for this thesis.

The framework for the proposed experiments currently exists. This framework relies on ETOL’s path-planning capabilities. It relies on CFD-based wind models that account for building. It relies on an HPC infrastructure that parallelizes simulated vehicles across computational nodes, and this HPC feature has been demonstrated with a Georgia Tech Partnership for an Advanced Computing Environment (PACE) HPC cluster. With this framework, this thesis will have reproducible results because the framework encases experiments in Singularity containers.

Committee

• Prof. Eric Feron – School of Aerospace Engineering (advisor)
• Prof. Brian German – School of Aerospace Engineering (co-advisor)
• Prof. Graeme Kennedy – School of Aerospace Engineering